Professional Documents
Culture Documents
New Science
Other Books by Howard Gardner
HOWARD GARDNER
PREFACE .riii
PART I
THE COGNITIVE REVOLUTION
1 Introduction: What the Meno Wrought J
The Greek Agenda J
DefiniHon and Scope of CogniH'IJI! Science 5
Purpose and Plan of This Book 7
vii
Contents
PART II
THE COGNITIVE SCIENCES: A HISTORICAL
PERSPECTIVE
viii
Contents
Frederic Bartlett's Schematic Approach 114
Jean Piaget 's Developmental Concerns 116
The Tum to Cognition 118
lnspiratitm from Computers 118
Reactions to the Standard Information-Processing Paradigms:
The Top-Down Perspective 124
Mental Representations 128
Psychology's Contributions 130
X
Contents
PART III
TOWARD AN INTEGRATED COGNITIVE
SCIENCE: PRESENT EFFORTS,
FUTURE PROSPECTS
Introduction 291
xi
Contents
Cards with Numbers 361
Artists and Beekeepers 363
Mental Models as a Panacea? 367
Biases in Human Cognition: The Tversky-Kahneman
Position 370
Theater Tickets and Coin Tossers 371
A Philosophical Critique 3 73
Conclusion 3 79
REFERENCES 393
NAME INDEX 409
SUBJECT INDEX 415
Xtt
PREFACE
xm
Preface
abroad. As far as I can recall, no scientist whom I approached denied me
an interview, and most--even those who expressed skepticism about cog-
nitive science-were gracious and informative. I regret that I had to stop
interviewing and begin writing after a time, and I regret even more that
I ultimately was not able to discuss in print the work of many of those from
whom I learned much. Unfortunately, if I had included even half of the
work worthy of review, this book would be several times longer than
it is.
I want to mention first and thank the many individuals who willingly
discussed their work and the field of cognitive science with me. (I also must
apologize to those whom I have inadvertently omitted from this list.) I am
indebted to Jonathan Adler, Allan Allport, John Anderson, Dana Ballard,
Jon Barwise, Elizabeth Bates, Brent Berlin, Ned Block, Daniel Bobrow,
Margaret Boden, Stanley Brandes, Joan Bresnan, John Seely Brown, Roger
Brown, Jerome Bruner, Peter Bryant, Alfonso Caramazza, Noam Chomsky,
Gillian Cohen, Michael Cole, Roy D'Andrade, Daniel Dennett, Hubert
Dreyfus, Jerome Feldman, Charles Fillmore, Jerry Fodor, Michael Gaz-
zaniga, Clifford Geertz, my late and beloved mentor Norman Geschwind,
Samuel Glucksberg, Nelson Goodman, Charles Gross, Patrick Hayes,
Geoffrey Hinton, Stephen Isard, Philip Johnson-Laird, Ronald Kaplan,
Paul Kay, Samuel Jay Keyser, Stephen Kosslyn, George Lakoff, Jean Lave,
Jerome Lettvin, Robert LeVine, Claude Levi-Strauss, Christopher Longuet-
Higgins, John McCarthy, Jay McClelland, Jean Mandler, Alexander Mar-
shack, John Marshall, Jacques Mehler, Susanna Millar, George Miller,
Marvin Minsky, Julius Moravcsik, John Morton, Ulric Neisser, Freda
Newcombe, Allen Newell, Donald Norman, Daniel Osherson, Domenico
Parisi, Stanley Peters, Michael Posner, Karl Pribram, Hilary Putnam, Raj
Reddy, Richard Rorty, Eleanor Rosch, David Rumelhart, Roger Schank,
Israel Scheffler, John Searle, Robert Siegler, Herbert Simon, Aaron Sloman,
Brian Cantwell Smith, Stuart Sutherland, Leonard Talmy, Sheldon
Wagner, Terry Winograd, and Edgar Zurif.
Several friends and colleagues were good enough to read and comment
critically on one or more of the drafts of this book. I am considerably in
their debt. I wish to thank Margaret Boden, Hiram Brownell, Daniel Den-
nett, Martha Farah, Josef Grodzinsky, Jerome Kagan, Benny Shanon, Eric
Wanner, my wife, Ellen Winner, and several anonymous reviewers for
their useful comments, criticisms, and words of encouragement. I know
that I benefited greatly from their feedback; I fear that remaining errors and
infelicities are my own responsibility.
Over the several years in which this book was in preparation, I was
fortunate enough to have the help of Linda Levine, Susan McConnell,
Christine Meyer, and Claudia Strauss, who served as research assistants.
xiv
Preface
HowARD GARDNER
Cambridge, Massachusetts
Apri/1985
XV
PART I
THE COGNITIVE
REVOLUTION
1
Introduction:
What the Meno
Wrought
One thing I would fight for to the end, both in word and
deed if I were able-that if we believed that we must try
to find out what is not known, we should be better and
braver and less idle than if we believed that what we do
not know it is impossible to find out and that we need not
even try.
-SOCRATES, The Meno
3
I I THE COGNITIVE REVOLUTION
4
Introduction: What the Meno Wrought
five hundred years ago. Like their earlier counterparts, cognitive scientists
today ask what it means to know something and to have accurate beliefs,
or to be ignorant or mistaken. They seek to understand what is known-
the objects and subjects in the external world-and the person who knows
-his" perceptual apparatus, mechanisms of learning, memory, and ratio-
nality. They ponder the sources of knowledge: where does it come from,
how is it stored and tapped, how might it be lost? They are curious about
the differences among individuals: who learns early or with difficulty;
what can be known by the child, the inhabitant of a preliterate society, an
individual who has suffered brain damage, or a mature scientist?
Further, cognitive scientists, again as did the Greeks, conjecture about
the various vehicles of knowledge: what is a form, an image, a concept, a
11
word; and how do these modes of representation" relate to one another?
They wonder about the priorities of specific sense organs as against a
central"general understanding" or "common sense." They reflect on lan-
guage, noting the power and traps entailed in the use of words and their
possible predominant influence over thoughts and beliefs. And they spec-
ulate at length on the nature of the very activity of knowing: why do we
want to know, what are the constraints on knowing, and what are the
limits of scientific knowledge about human knowing?
This "new science," thus, reaches back to the Greeks in the commit-
ment of its members to unraveling the nature of human knowledge. At the
same time, however, it is radically new. Proceeding well beyond armchair
speculation, cognitive scientists are fully wedded to the use of empirical
methods for testing their theories and their hypotheses, of making them
susceptible to disconfirmation. Their guiding questions are not just a re-
hash of the Greek agenda: new disciplines, like artificial intelligence, have
arisen; and new questions, like the potential of man-made devices to think,
stimulate research. Moreover, cognitive scientists embrace the most recent
scientific and technological breakthroughs in a variety of disciplines. Most
central to their undertaking is the computer-that creation of the mid-
twentieth century that holds promise for changing our conceptions of the
world in which we live and our picture of the human mind.
5
I I THE COGNITIVE REVOLUTION
field (see, for example, Bruner 1983; Collins 1977; Mandler 1981; Miller
1979; Norman 1980; Rumelhart 1982). It therefore becomes important for
me at the outset to state what I take cognitive science to be.
I define cognitive science as a contemporary, empirically based effort
to answer long-standing epistemological questions-particularly those
concerned with the nature of knowledge, its components, its sources, its
development, and its deployment. Though the term cognilive sdence is some-
times extended to include all forms of knowledge-animate as well as
inanimate, human as well as nonhuman-! apply the term chiefly to efforts
to explain human knowledge. I am interested in whether questions that
intrigued our philosophical ancestors can be decisively answered, instruc-
tively reformulated, or permanently scuttled. Today cognitive science
holds the key to whether they can be.
Of the various features or aspects generally associated with cognitive-
scientific efforts, I consider five to be of paramount importance. Not every
cognitive scientist embraces every feature, of course; but these features can
be considered symptomatic of the cognitive-scientific enterprise. When all
or most are present, one can assume that one is dealing with cognitive
science; when few, if any, are present, one has fallen outside my definition
of cognitive science. These features will be introduced more formally at the
end of chapter 3 and will be revisited repeatedly throughout the book, but
it is important to make an initial acquaintance with them at this point.
First of all, there is the belief that, in talking about human cognitive
activities, it is necessary to speak about mental representations and to posit
a level of analysis wholly separate from the biological or neurological, on
the one hand, and the sociological or cultural, on the other.
Second, there is the faith that central to any understanding of the
human mind is the electronic computer. Not only are computers indispens-
able for carrying out studies of various sorts, but, more crucially, the
computer also serves as the most viable model of how the human mind
functions.
While the first two features incorporate the central beliefs of current
cognitive science, the latter three concern methodological or strategic char-
acteristics. The third feature of cognitive science is the deliberate decision
to de-emphasize certain factors which may be important for cognitive
functioning but whose inclusion at this point would unnecessarily compli-
cate the cognitive-scientific enterprise. These factors include the influence
of affective factors or emotions, the contribution of historical and cultural
factors, and the role of the background context in which particular actions
or thoughts occur.
As a fourth feature, cognitive scientists harbor the faith that much is
to be gained from interdisciplinary studies. At present most cognitive
6
Introduction: What the Meno Wrought
scientists are drawn from the ranks of specific disciplines-in particular,
philosophy, psychology, artificial intelligence, linguistics, anthropology,
and neuroscience (I shall refer to these disciplines severally as the "cogni-
tive sciences"). The hope is that some day the boundaries between these
disciplines may become attenuated or perhaps disappear altogether, yield-
ing a single, unified cognitive science.
A fifth and somewhat more controversial feature is the claim that a
key ingredient in contemporary cognitive science is the agenda of issues,
and the set of concerns, which have long exercised epistemologists in the
Western philosophical tradition. To my mind, it is virtually unthinkable
that cognitive science would exist, let alone assume its current form, had
there not been a philosophical tradition dating back to the time of the
Greeks.
7
I 1 THE CoGNITIVE REvoLUTION
have recurred and the kinds of approaches that are especially central
within a particular field. For example, in philosophy I trace the perennial
dispute between those of a rationalist persuasion (who view the mind as
actively organizing experiences on the basis of pre-existing schemes); and
those of an empiricist bent (who treat mental processes as a reflection of
information obtained from the environment). In anthropology I survey
various attempts over the years to compare the thought of primitive peo-
ples with that exhibited by typical individuals in modem Western society.
Approaching these same fields from a methodological point of view, I raise
the questions whether philosophy will eventually come to be supplanted
by an empirically based cognitive science, and whether anthropology can
(or even should) ever transcend the individual case study.
Of course, such organizing themes can only scratch the surface of the
complex territory that underlies any scientific discipline. Still I hope that
through such themes I can convey how a linguist views an issue, what a
psychologist deems a problem (and a solution), which conceptions of
process obtain in neuroscience and artificial intelligence. Only through
such an immersion in the daily (and yearly) concerns of a cognitive scien-
tist drawn from a particular discipline can one appreciate the possibilities
-and the difficulties-that arise when workers from different fields col-
laborate in cognitive-scientific research. In the end I will in each case take
stock and indicate where things stand with reference to the principal lines
of contention in a particular cognitive science-a discussion that will, in
tum, suggest some of the principal factors that have stimulated cognitive
scientists to join forces.
While each of the histories stands on its own, their juxtaposition
points up fascinating and difficult-to-anticipate parallels. Scientific fields
hardly develop in a vacuum: such disparate factors as the dissemination
of Darwin's pivotal writings, the outbreak of wars, the rise of great univer-
sities have had reverberations-and sometimes cataclysmic ones-across
apparently remote fields, which may well have had little direct contact
with one another. For the most part, I shall simply allow these parallels to
emerge, but at the beginning of part III I shall specify certain historical
forces that seem to have exerted influence across a range of cognitive
sciences.
Having taken the measure of the individual cognitive sciences, I tum
in the third part of the book to review ongoing work that is quintessen-
tially cognitive-scientific. Thus, in chapters 10 to 13, the focus shifts from
work within a traditional discipline to those lines of research that stand
most squarely at the intersection of a number of disciplines and therefore
can be considered prototypical of a single, unified cognitive science. I have
sought to identify work that is of the highest quality: if cognitive science
8
Introduction: What the Meno Wrought
is to be assessed as an intellectual enterprise, it ought to be judged by the
most outstanding instances.
There is a common structure to these four essays on current cognitive-
scientific work. Consistent with my claim that cognitive science seeks to
elucidate basic philosophical questions, each chapter begins with a peren-
nial epistemological issue. For example, in chapter 10, I describe work on
how we perceive the world; in chapter 13, I review competing claims on
the extent of human rationality. Across chapters 10 to 13, there is a pro-
gression from those issues that seem most circumscribed to those that are
most global. Not surprisingly, the most confident answers exist for the
delimited questions, while the global topics remain ringed by unresolved
questions.
My personal reflections on cognitive science are reserved for the final
chapter. There I revisit the major themes of cognitive science in light of the
histories sketched and the interdisciplinary work reviewed. I also discuss
two themes that emerge from the inquiry and that will be introduced at
greater length in chapter 3: the computational paradox and the cognitive
challenge. In my view, the future of cognitive science rests on how the
computational paradox is resolved and on how the cognitive challenge is
met.
One might say that cognitive science has a very long past but a
relatively short history. The reason is that its roots go back to classical
times, but it has emerged as a recognized pursuit only in the last few
decades. Indeed, it seems fair to maintain that the various components that
gave rise to cognitive science were all present in the early part of the
century, and the actual birthdate occurred shortly after mid-century. Just
why cognitive science arose when it did in the form it did will constitute
my story in the remainder of part I.
9
2
Laying the
Foundation for
Cognitive Science
10
Laying the Foundation for Cognitive Science
chologist Karl Lashley, gave the most iconoclastic and most memorable
address. Speaking on "The Problem of Serial Order in Behavior," he chal-
lenged the doctrine (or dogma) that had dominated psychological analysis
for the past several decades and laid out a whole new agenda for research.
In the terms of my own discussion, Lashley identified some of the major
components needed for a cognitive science, even as he castigated those
forces that had prevented its emergence before this time.
In order to appreciate the importance of Lashley's remarks, it is neces-
sary to consider the scientific climate in which he (and his numerous
colleagues interested in human psychology) had been operating during the
past few decades. At the turn of the century, in the wake of the founding
of new human sciences, investigators had been addressing the key issues
of mental life: thinking, problem solving, the nature of consciousness, the
unique aspects of human language and culture. These discussions had
linked up with the philosophical agenda of the West, but investigators had
sought to go beyond sheer speculation through the use of rigorous experi-
mental methods.
Unfortunately the scientific method favored by most researchers at
that time was introspection: self-reflection on the part of a trained observer
about the nature and course of his own thought patterns. Though sugges-
tive (indeed, often too suggestive), such introspection did not lead to that
accumulation of knowledge that is critical to science. lntrospectionism
might have collapsed of its own weight, but, in fact, it was toppled in a
more aggressive manner by a group of mostly young, mostly American
scientists who became known as the "behaviorists."
The behaviorists put forth two related propositions. First of all, those
researchers interested in a science of behavior ought to restrict themselves
strictly to public methods of observations, which any scientist could apply
and quantify. No subjective ruminations or private introspection: if a
discipline were to be science, its elements should be as observable as the
physicist's cloud chamber or the chemist's flask. Second, those interested
in a science of behavior ought to focus exclusively on behavior: researchers
ought assiduously to eschew such topics as mind, thinking, or imagination
and such concepts as plans, desires, or intentions. Nor ought they to
countenance hypothetical mental constructs like symbols, ideas, schemas,
or other possible forms of mental representation. Such constructs, never
adequately clarified by earlier philosophers, had gotten the introspection-
ists into hot water. According to behaviorists, all psychological activity can
be adequately explained without resorting to these mysterious mentalistic
entities.
A strong component of the behaviorist canon was the belief in the
supremacy and determining power of the environment. Rather than in-
11
I I THE CoGNITIVE REVOLUTION
12
Laying the Foundation for Cognitive Science
of human psychology a topic that had been relegated to obscurity by his
behaviorist colleagues. At the same time, he added, the dominant theoreti-
cal explanatory framework in neurophysiology no less than in psychology
-that of simple associative chains between a stimulus and a response-
could not possibly account for any of this serially ordered behavior. The
reason is that these action sequences unfold with such rapidity that there
is no way in which the next step in the chain can be based upon the
previous one: when one plays an arpeggio, for instance, there is simply no
time for feedback, no time for the next tone to depend upon or in any way
to reflect the course of the preceding one. Similarly, the kinds of error made
by individuals-for example, slips of the tongue-often include anticipa-
tion of words that are to occur only much later in a sequence. Again, these
phenomena defy explanations in terms of linear "A evokes B" chains.
According to Lashley, these behavioral sequences have to be planned
and organized in advance. The organization is best thought of as hierarchi-
cal: there are the broadest overall plans, within which increasingly fine-
grained sequences of actions are orchestrated. Thus, for instance, in the
case of speech, the highest nodes of the hierarchy involve the overall
intention prompting the utterance, while the choice of syntax and the
actual production of sounds occupy lower nodes of the hierarchy. The
nervous system contains an overall plan or structure, within which indi-
vidual response units can-indeed, have to-be slotted, independent of
specific feedback from the environment. Rather than behavior being con-
sequent upon environmental promptings, central brain processes actually
precede and dictate the ways in which an organism carries out complex
behavior. Or, to put it simply, Lashley concluded that the form precedes
and determines specific behavior: rather than being imposed from without,
organization emanates from within the organism.
Even as he defied the standard behavioral analysis of the time, Lashley
was also challenging two major dogmas of neurobehavioral analysis: the
belief that the nervous system is in a state of inactivity most of the time,
and the belief that isolated reflexes are activated only when specific forms
of stimulation make their appearance. Lashley's nervous system consisted
of always active, hierarchically organized units, with control emanating
from the center rather than from peripheral stimulation. As he put it,
"Attempts to express cerebral function in terms of the concepts of the
reflex arc, or of associated chains of neurons, seem to me doomed to failure
because they start with the assumption of a static nervous system. Every
bit of evidence available indicated a dynamic, constantly active system, or,
rather, a composite of many interacting systems" (quoted in Jeffress 1951,
p. 135).
In the topics he chose to address, and in the ways in which he ad-
13
I I THE CoGNITIVE REvOLUTION
14
Laying the FoundaHon for CogniHve Science
force), these scholars at the end of the nineteenth century had launched
separate scientific disciplines, like psychology, linguistics, anthropology,
sociology, and various neurosciences. That these aspiring scientists of
human nature had succeeded in establishing effective institutional bases
within the universities could not be disputed; but the extent to which each
new discipline had arrived at important truths was still being debated at
mid-century. Finally, those attending the Pasadena meeting were well
acquainted with the scientific program of the behaviorists. And they
shared an intuition-strongly bolstered by Lashley's tightly reasoned
paper-that the behaviorist answer to questions of the human mind was
no answer at all.
But other factors had also impeded the proper launching of a science
of cognition. Fitting comfortably with behaviorism were several philo-
sophical schools-positivism, physicalism, and verificationism-which es-
chewed entities (like concepts or ideas) that could not be readily observed
and reliably measured. There was also the intoxication with psychoanal-
ysis. While many scholars were intrigued by Freud's intuitions, they felt
that no scientific discipline could be constructed on the basis of clinical
interviews and retrospectively constructed personal histories; moreover,
they deeply resented the pretense of a field that did not leave itself suscep-
tible to disconfirmation. Between the "hard line" credo of the Establish-
ment behaviorists and the unbridled conjecturing of the Freudians, it was
difficult to focus in a scientifically respectable way on the territory of
human thought processes.
Finally, the world political situation had exerted a crippling effect on
the scientific enterprise. First, the European scientific establishment had
been ripped apart by the rise of totalitarianism, and then the American
scientific establishment had been asked to lay aside its theoretical agenda
in order to help wage the war.
While the war had been, in many ways, the worst of times, bringing
on the death or disability of many talented investigators, it had also stimu-
lated certain scientific and technological activities. Within the United
States, the war effort demanded calculating machines that could "crunch"
large sets of numbers very quickly. Computers soon became a reality.
There were other war needs to be met as well. For instance, the mathemati-
cian Norbert Wiener was asked to devise more accurate anti-aircraft ma-
chinery. This work required "a good gun, a good projectile, and a fire-
control system that enables the gunner to know the target's position, apply
corrections to the gun controls, and set the fuse properly, so that it will
detonate the projectile at the right instant" (quoted in Heims 1980, p. 183).
While working on these problems at the Massachusetts Institute of Tech-
nology, Wiener and his associate, a young engineer named Julian Bigelow,
15
I I THE CoGNITIVE REvOLUTION
concluded that there were important analogies between the feedback as-
pects of engineering devices and the homeostatic processes by which the
human nervous system sustains purposive activity. These ideas of plan-
ning, purpose, and feedback, developed with mathematical precision, were
directly antithetical to the behaviorist credo. War also produced many
victims of gunfire; and medical practitioners who cared for brain-injured
patients were being asked to evaluate which tasks could be carried out and
which ones had been compromised-temporarily or permanently-by in-
jury to the nervous system. Also, a host of more person-centered issues-
ranging from the study of the effects of propaganda to the selection of men
fit to lead combat units-enlisted the efforts of behavioral scientists and
generated ideas on which the postwar human sciences were to build
(Bruner 1944; Murray 1945; Stouffer et al. 1949). So it was in other war-
tom lands, from Alan Turing and Kenneth Craik's interest in computers
in England, to Alexander Luria's painstaking research with brain-injured
patients in Russia during the war.
By the late 1940s, there was beginning to be a feeling abroad-one
observable at Pasadena but in no way restricted to that site-that perhaps
the time was ripe for a new and finally effective scientific onslaught on the
human mind. Interestingly, nearly all of the work that came to fruition in
the postwar era was in fact built upon prior theoretical efforts-work often
dating back to the beginning of the century. But this work had sometimes
been obscured by the behaviorist movement and had sometimes been
transformed in unanticipated ways by the events of the war. These ideas,
these key inputs to contemporary efforts in cognitive science, were already
familiar to the participants at the Hixon Symposium and to other scholars
involved in the first concerted efforts to found cognitive science during the
1940s and 1950s. Now it was time to put these ideas to optimal scientific
use.
16
Laying the Foundation for Cognitive Science
to evolve toward the end of the nineteenth century. Then, in the early
twentieth century, as I shall elaborate in chapter 4, the British mathemati-
cal logicians Bertrand Russell and Alfred North Whitehead sought, with
considerable success, to reduce the basic laws of arithmetic to propositions
of elementary logic. The Whitehead-Russell work influenced a whole gen-
eration of mathematically oriented thinkers, including both Norbert Wie-
ner and John von Neumann, two of the most important contributors to the
founding of cognitive science.
In the 1930s, the logical-mathematical work that ultimately had the
greatest import for cognitive science was being carried out by a then
relatively unknown British mathematician, Alan Turing. In 1936, he devel-
oped the notion of a simple machine (subsequently dubbed a "Turing
machine") which could in principle carry out any possible conceivable
calculation. The notions underlying this "theoretical" machine were sim-
ple. All one needed was an infinitely long tape which could pass through
the machine and a scanner to read what was on the tape. The tape itself
was divided into identical squares, each of which contained upon it either
a blank or some kind of slash. The machine could carry out four moves
with the tape: move to the right, move to the left, erase the slash, or print
the slash. With just these simple operations, the machine could execute
any kind of program or plan that could be expressed in a binary code (for
example, a code of blanks and slashes). More generally, so long as one
could express clearly the steps needed to carry out a task, it could be
programmed and carried out by the Turing machine, which would simply
scan the tape (no matter what its length) and carry out the instructions
(Davis 1958; McCorduck 1979).
Turing's demonstration-and the theorem he proved-was of pro-
found importance for those researchers interested in computing devices. It
suggested that a binary code (composed simply of zeros and ones) would
make possible the devising and execution of an indefinite number of
programs, and that machines operating on this principle could be con-
structed. As Turing himself pondered computing devices, he became in-
creasingly enthusiastic about their possibilities. In fact, in 1950 (shortly
before his untimely death by suicide in his early forties) he suggested that
one could so program a machine that it would be impossible to discrimi-
nate its answers to an interlocutor from those contrived by a living human
being-a notion immortalized as the "Turing machine test." This test is
used to refute anyone who doubts that a computer can really think: if an
observer cannot distinguish the responses of a programmed machine from
those of a human being, the machine is said to have passed the Turing test
(Turing 1963).
The implications of these ideas were quickly seized upon by scientists
17
I I THE CoGNITIVE REVOLUTION
interested in human thought, who realized that if they could describe with
precision the behavior or thought processes of an organism, they might be
able to design a computing machine that operated in identical fashion. It
thus might be possible to test on a computer the plausibility of notions
about how a human being actually functions, and perhaps even to con-
struct machines about which one could confidently assert that they think
just like human beings.
In building upon Turing's ideas, John von Neumann pursued the
notion of devising a program to instruct the Turing machine to reproduce
itself. Here was the powerful idea of a stored program: that is, the computer
could be controlled through a program that itself was stored within the
computer's internal memory, so that the machine would not have to be
laboriously reprogrammed for each new task (see Goldstine 1972). For the
first time, it became conceivable that a computer might prepare and
execute its own programs.
18
Laying the Foundation {or Cognitive Science
a computer that could realize any process that can be unambiguously
described. Turing had demonstrated the possibility in principle of comput-
ing machines of great power, while McCulloch and Pitts had demonstrated
that at least one redoubtable machine-the human brain-could be
thought of as operating via the principles of logic and, thus, as a powerful
computer.
Ultimately, McCulloch may have carried his own chain of reasoning
too far. He was convinced that fundamental problems of epistemology
could be stated and solved only in light of the knowledge of the central
nervous system (McCorduck 1979), and he tied his claims about thinking
very closely to what was known during his own time about the nervous
system. Some commentators even feel that the search by McCulloch and
his associates for a direct mapping between logic machines and the nervous
system was a regressive element in the development of cognitive science:
rather than trying to build machines that mimic the brain at a physiological
level, analogies should have been propounded and pursued on a much
higher level-for example, between the thinlcing that goes on in human
problem solving and the strategies embodied in a computer program
(McCarthy 1984). On the other hand, it was due in part to McCulloch's
own analysis that some of the most important aspects of the nervous
system came to be better understood: for he sponsored research on the
highly specific properties of individual nerve cells. Moreover, very recently
computer scientists have once again been drawing directly on ideas about
the nature of and connections among nerve cells (see chapter 10, pp.
318-22). On balance, his polymathic spirit seems to have been a benign
catalyst for the growth of cognitive science.
siles, and airplanes on course-he had come to think about the nature of
feedback and of self-correcting and self-regulating systems, be they me-
chanical or human. He collaborated closely with Vannevar Bush, who had
pioneered in the development of analog computers. Wiener was also struck
by the importance of the work of his sometime colleagues McCulloch and
Pitts, particularly by the suggestive analogies between a system of logical
connections and the human nervous system.
Wiener went beyond all of his contemporaries in his missionary con-
viction that these various scientific and technological developments co-
hered. Indeed, in his mind they constituted a new science-one founded
on the issues of control and communication, which he deemed to be central
in the middle of the twentieth century. He first publicly formulated this
point of view in a 1943 paper, "Behavior, Purpose, and Teleology" (Ro-
senblueth, Wiener, and Bigelow 1943), in which he and his fellow authors
put forth the notion that problems of control engineering and communica-
tion engineering are inseparable; moreover, they center not on the tech-
niques of electrical engineering, but rather on the much more fundamental
notion of the message-"whether this should be transmitted by electrical,
mechanical, or nervous means." The authors introduced a then-radical
notion: that it is legitimate to speak of machines that exhibit feedback as
"striving toward goals," as calculating the difference between their goals
and their actual performance, and as then working to reduce those differ-
ences. Machines were purposeful. The authors also developed a novel
notion of the central nervous system. As Wiener later put it:
20
Laying the FoundaHon for CogniHve Science
set down an integrated vision-a linkage of developments in understand-
ing the human nervous system, the electronic computer, and the operation
of other machines. And he underscored his belief-echoing von Neumann
and McCulloch and Pitts-that the functioning of the living organism and
the operation of the new communication machines exhibited crucial paral-
lels. Though Wiener's synthesis was not ultimately the one embraced by
cognitive science (it came closer to achieving that exalted status in the
Soviet Union), it stands as a pioneering example of the viability of such
an interdisciplinary undertaking.
Information Theory
Another key progenitor of cognitive science was Claude Shannon, an
electrical engineer at M.I.T. who is usually credited with devising informa-
tion theory. Already as a graduate student at M.I.T. in the late 1930s,
Shannon had arrived at a seminal insight. He saw that the principles of
logic (in terms of true and false propositions) can be used to describe the
two states (on and off) of electromechanical relay switches. In his master's
thesis, Shannon provided an early suggestion that electrical circuits (of the
kind in a computer) could embody fundamental operations of thought. I
shall describe this work-so crucial for all subsequent work with comput-
ers-further in chapter 6.
During the next ten years, working in part with Warren Weaver,
Shannon went on to develop the key notion of information theory: that
information can be thought of in a way entirely divorced from specific
content or subject matter as simply a single decision between two equally
plausible alternatives. The basic unit of information is the bit (short for
"binary digit"): that is, the amount of information required to select one
message from two equally probable alternatives. Thus, the choice of a
message from among eight equally probable alternatives required three
bits of information: the first bit narrowed the choice from one of eight to
one of four; the second, from one of four to one of two; the third selects
one of the remaining alternatives. Wiener explained the importance of this
way of conceptualization: "Information is information, not matter or en-
ergy. No materialism which does not admit this can survive at the present
day" (Wiener 1961, p. 132).
Thanks to Wiener's insights, it became possible to think of informa-
tion apart from a particular transmission device: one could focus instead
on the efficacy of any communication of messages via any mechanism, and
one could consider cognitive processes apart from any particular embodi-
ment-an opportunity upon which psychologists would soon seize as they
sought to describe the mechanisms underlying the processing of any kind
21
I I THE CoGNITIVE REvoLUTION
Neuropsychological Syndromes
A comparable contribution to an incipient cognitive science came
from a remote and unexpected scientific comer-the profiles of cognitive
incapacities following damage to the human brain. Paradoxically, this area
of science relies heavily on the travesties of war. As in the era of the First
World War, much was learned during the Second World War about
aphasia (language deficit), agnosia (difficulty in recognition), and other
forms of mental pathology consequent upon injury to the brain. Laborato-
ries in New York, Oxford, Paris, Berlin, and Moscow were all busily
engaged in working with victims of brain damage. When the neuropsycho-
logical researchers began to communicate their findings to one another,
considerable convergence was noted even across cultural and linguistic
boundaries. For instance, aphasia assumed similar forms despite wide dif-
ferences across languages. There was, it seemed, much more regularity in
the organization of cognitive capacities in the nervous system than was
allowed for by wholly environmental accounts of mental processes. Fur-
thermore, the patterns of breakdown could not be readily explained in
terms of simple stimulus-response disruption. Rather, in many cases, the
hierarchy of behavioral responses was altered. For example, in certain
forms of aphasia, the general sentence frame was preserved, but subjects
could not correctly slot individual words into the frame. In other aphasias,
the sentence frame broke down, but individual content words carried
meaning. Thus was struck yet another blow against reflex-arc models of
thought. At the same time, the particular profiles of abilities and disabili-
ties that emerge in the wake of brain damage provided many pregnant
suggestions about how the human mind might be organized in normal
individuals.
By the late 1940s, in areas as diverse as communication engineering
and neuropsychology, certain themes were emerging principally in the
United States, Great Britain, and the Soviet Union. Though I have stressed
the American version of this story, comparable accounts could be pre-
sented from other national perspectives as well. Scholars in these fields
were not only writing but were eagerly meeting with one another to
discuss the many exciting new perspectives. Herbert Simon, ultimately one
of the founders of cognitive science but then a graduate student at the
University of Chicago, recalls a kind of "invisible college" in the 1940s
(Simon 1982). He knew McCulloch at Chicago; he knew of Shannon's
master's thesis at M.I.T.; he knew that Wiener and von Neumann were
22
Laying the Foundation for Cognitive Science
working on issues in symbolic logic which had grown out of the philo-
sophical writings of Whitehead, Russell, and Frege. Simon himself was
studying at Chicago with Rudolf Camap, who was then putting forth key
notions about the syntax of logic. Such leading biologists (and Hixon
symposiasts) as Ralph Gerard, Heinrich Kluver, Roger Sperry, and Paul
Weiss were working in nearby laboratories on issues of the nervous sys-
tem. Many of the same influences were rubbing off during this period on
Jerome Bruner, Noam Chomsky, John McCarthy, George Miller, Allen
Newell, and other founders of cognitive science.
23
I I THE CoGNITIVE REVOLUTION
Foundation looms large. In the winter of 1944, John von Neumann and
Norbert Wiener convened a meeting at Princeton of all those interested in
what later came to be called "cybernetics." Present at the Macy-sponsored
event were many of the scholars already introduced in this narrative.
Wiener later recalled, "At the end of the meeting it had become clear to
all that there was a substantial common basis of ideas between the workers
in the different fields, that people in each group could already use notions
which had been better developed by the others, and that some attempt
should be made to achieve a common vocabulary" (1961, p. 15).
Building on these initial contacts, Warren McCulloch arranged with
the Macy Foundation in the spring of 1946 for a series of meetings on the
problems of feedback. "The idea has been to get together a group of
modest size, not exceeding some twenty in number, of workers in various
related fields and to hold them together for two successive days in aU-day
series of informal papers, discussions, and meals together, until they had
had the opportunity to thresh out their differences and to make progress
in thinking along the same lines" (Wiener 1961, p. 18). Ultimately there
were ten such meetings, about one a year, of what was originally the
Conference for Circular Causal and Feedback Mechanisms in Biological
and Social Systems-soon (and happily) shortened, at Wiener's urging, to
the Conference on Cybernetics. In the transcripts of these conferences, one
discerns ample evidence of scholars informing one another as well as first
intimations of interesting and sometimes unexpected projects. For exam-
ple, it was in discussions at the Macy meetings that the anthropologist
Gregory Bateson first encountered ideas about feedback which he was to
mine in his "double-bind" theory of schizophrenia.
Activity was especially intense in the Boston and Princeton areas and
in California. During the early 1950s, J. Robert Oppenheimer, director of
the Princeton Institute for Advanced Study (of which von Neumann was
a permanent member) became interested in the application of some of
these new ideas in the field of psychology. He regularly invited a group
of psychologists to visit the institute and report on recent developments
in their field. Among those who spent a year there were George Miller and
Jerome Bruner, gifted young psychologists who would shortly play a
fundamental role in the launching of cognitive science.
Again, there was difficult-to-anticipate but promising cross-fertiliza-
tion of ideas. Oppenheimer was particularly interested in analogies be-
tween the problems of perception, as they are viewed by the psychologist,
and issues of observation, which had come to loom large in atomic and
subatomic physics, once one began to work at the atomic and the sub-
atomic levels. He had been pondering the disturbing implications of the
indeterminacy principle, according to which it is impossible to ascertain the
24
Laying the Foundation for Cognitive Science
position and the velocity of a particle without affecting it during the course
of measurement. Meanwhile, Bruner had been studying the effects of an
observer's attitude and expectations on putatively 0bjective data." One
11
11
day Oppenheimer remarked to him, Perception as you psychologists
study it can't, after all, be different from observation in physics, can it?"
(quoted in Bruner 1983, pp. 95-96).
In Boston, discussion of these cognitive themes was continuing at
M.I. T ., and at the associated Lincoln Laboratories, where a group of young
engineers and psychologists had assembled to work on applied problems,
such as early warning signals in the case of bomb attacks. At nearby
Harvard in the prestigious Society of Fellows, the influence of behaviorist
thinking was dominant among the senior fellows, but the young junior
fellows, including the linguist Noam Chomsky and the mathematician
Marvin Minsky, were already proceeding in different (and anti-behavio-
rist) theoretical directions (Miller 1982). The Ford Foundation, having
decided to help stimulate work in the behavioral sciences, established a
Center for Advanced Study in the Behavioral Sciences in Palo Alto and
also provided funding for a significant proportion (perhaps one third) of
all the research psychologists in America. At the Rand Corporation in
Southern California, groups of mathematicians and engineers were work-
ing on the development of computing machines. Two young scientists,
Allen Newell and Herbert Simon, had begun to talk about the possibilities
of creating machines that could genuinely think. And, again, there was a
British version as well-the Ratio Club, which commenced in 1949. Cen-
tral to the Ratio Club was the notion of processing information in animals
and machines. Members included physiologists, engineers, physicians, and
psychologists with interests in the mind or "minding." Turing occasionally
attended meetings. The group (which met for several years) had the in-
triguing rule that any member who reached the rank of full professor must
be expelled, because he would then have potential control over other
members (McCorduck 1979, p. 78).
25
I I THE CoGNITIVE REvOLUTION
26
Laying the Foundation for Cognitive Science
other institutions and venues. Important papers and books were being
written and discussed. Still, all of this activity was going on, in a sense,
outside established fields of study. It was extracurricular and considered
a bit odd by those in the mainstream-behaviorist psychology, structural
linguistics, functionalist social anthropology, the neuropsychology of ani-
mal learning. It would take more dramatic events to shake these fields to
their foundation-events that were not long in coming.
27
3
Cognitive Science:
The First Decades
A Consensual Birthdate
Seldom have amateur historians achieved such consensus. There has been
nearly unanimous agreement among the surviving principals that cognitive
science was officially recognized around 1956. The psychologist George A.
Miller (1979) has even fixed the date, 11 September 1956.
Why this date? Miller focuses on the Symposium on Information
Theory held at the Massachusetts Institute of Technology on 10-12 Sep-
tember 1956 and attended by many leading figures in the communication
and the human sciences. The second day stands out in Miller's mind
because of two featured papers. The first, presented by Allen Newell and
Herbert Simon, described the "Logic Theory Machine," the first complete
proof of a theorem ever carried out on a computing machine. The second
paper, by the young linguist Noam Chomsky, outlined "Three Models of
Language." Chomsky showed that a model of language production derived
from Oaude Shannon's information-theoretical approach could not possi-
bly be applied successfully to "natural language," and went on to exhibit
his own approach to grammar, based on linguistic transformations. As
Miller recalls, "Other linguists had said language has all the formal preci-
sions of mathematics, but Chomsky was the first linguist to make good on
the claim. I think that was what excited all of us" (1979, p. 8). Not inciden-
tally, that day George Miller also delivered a seminal paper, outlining his
claim that the capacity of human short-term memory is limited to approxi-
mately seven entries. Miller summed up his reactions:
28
Cognitive Science: The First Decades
I went away from the Symposium with a strong conviction, more intuitive
than rationat that human experimental psychology, theoretical linguistics, and
computer simulation of cognitive processes were all pieces of a larger whole, and
that the future would see progressive elaboration and coordination of their shared
concerns .... I have been working toward a cognitive science for about twenty years
beginning before I knew what to call it. (1979, p. 9)
For reasons-that are obscure at present, the various tensions and inadequacies
of the first half of the twentieth century cooperated to produce a new movement
in psychology that first adopted the label of information processing and after
became known as modern cognitive psychology. And it all happened in the five
year period between 1955 and 1960. Cognitive science started during that five year
period, a happening that is just beginning to become obvious to its practitioners.
(1981, p. 9)
Finally, in their history of the period, computer scientists Allen Newell and
Herbert Simon declare:
Within the last dozen years a general change in scientific outlook has oc-
curred, consonant with the point of view represented here. One can date the change
roughly from 1956: in psychology, by the appearance of Bruner, Goodnow, and
Austin's Study of Thinking and George Miller's "The magical number seven"; in
linguistics, by Noam Chomsky's "Three models of language"; and in computer
science, by our own paper on the Logical Theory Machine. (1972, p. 4)
29
I I THE CoGNITIVE REVOLUTION
ers and analyzed the idea of a program, the operation of memory in com-
puters, and the possibility of machines that replicate themselves.
Relevant research emanated from each of the fields that I have desig-
nated as contributing cognitive sciences.* The witnesses I have just quoted
noted the principal texts in the fields of psychology, linguistics, and artifi-
cial intelligence, and many more entries could be added. Neuroscientists
were beginning to record impulses from single neurons in the nervous
system. At M.I.T., Warren McCulloch's research team, led by the neuro-
physiologists Jerome Lettvin and Humberto Maturana, recorded from the
retina of the frog. They were able to show that neurons were responsive
to extremely specific forms of information such as "bug-like" small dark
spots which moved across their receptive fields, three to five degrees in
extent. Also in the late 1950s, a rival team of investigators, David Hubel
and Torsten Wiesel at Harvard, began to record from cells in the visual
cortex of the cat. They located nerve cells that responded to specific infor-
mation, including brightness, contrast, binocularity, and the orientation of
lines. These lines of research, eventually honored in 1981 by a Nobel Prize,
called attention to the extreme specificity encoded in the nervous system.
The mid 1950s were also special in the field of anthropology. At this
time, the first publications by Harold Conklin, Ward Goodenough, and
Floyd Lounsbury appeared in the newly emerging field of cognitive an-
thropology, or ethnosemantics. Researchers undertook systematic collec-
tion of data concerning the naming, classifying, and concept-forming abili-
ties of people living in remote cultures, and then sought to describe in
formal terms the nature of these linguistic and cognitive practices. These
studies documented the great variety of cognitive practices found around
the world, even as they strongly suggested that the relevant cognitive
processes are similar everywhere.
In addition, in the summer of 1956, a group of young scholars, trained
in mathematics and logic and interested in the problem-solving potentials
of computers, gathered at Dartmouth College to discuss their mutual inter-
ests. Present at Dartmouth were most of the scholars working in what
came to be termed "artificial intelligence," including the four men gener-
ally deemed to be its founding fathers: John McCarthy, Marvin Minsky,
Allen Newell, and Herbert Simon. During the summer institute, these
scientists, along with other leading investigators, reviewed ideas for pro-
grams that would solve problems, recognize patterns, play games, and
reason logically, and laid out the principal issues to be discussed in coming
years. Though no synthesis emerged from these discussions, the partici-
pants seem to have set up a permanent kind of "in group" centered at the
•Full bibliographical references to these lines of research will be provided at appropriate
points in the text.
30
Cognitive Science: The Rrst Decades
M.I.T., Stanford, and Carnegie-Mellon campuses. To artificial intelligence,
this session in the summer of 1956 was as central as the meeting at M.I.T.
among communication scientists a few months later.
Scholars removed from empirical science were also pondering the
implications of the new machines. Working at Princeton, the American
philosopher Hilary Putnam (1960) put forth an innovative set of notions.
As he described it, the development of Turing-machine notions and the
invention of the computer helped to solve-or to dissolve-the classical
mind-body problem. It was apparent that different programs, on the same
or on different computers, could carry out structurally identical problem-
solving operations. Thus, the logical operations themselves (the "soft-
ware") could be described quite apart from the particular "hardware" on
which they happened to be implemented. Put more technically, the
"logical description" of a Turing machine includes no specification of its
physical embodiment.
The analogy to the human system and to human thought processes
was clear. The human brain (or "bodily states") corresponded to the com-
putational hardware; patterns of thinking or problem solving ("mental
states") could be described entirely separately from the particular constitu-
tion of the human nervous system. Moreover, human beings, no less than
computers, harbored programs; and the same symbolic language could be
invoked to describe programs in both entities. Such notions not only
clarified the epistemological implications of the various demonstrations in
artificial intelligence; they also brought contemporary philosophy and em-
pirical work in the cognitive sciences into much closer contact.
One other significant line of work, falling outside cognitive science as
usually defined, is the ethological approach to animal behavior which had
evolved in Europe during the 1930s and 1940s thanks to the efforts of
Konrad Lorenz (1935) and Niko Tinbergen (1951). At the very time that
American comparative psychologists were adhering closely to controlled
laboratory settings, European ethologists had concluded that animals
should be studied in their natural habitat. Observing carefully under these
naturalistic conditions, and gradually performing informal experiments on
the spot, the ethologists revealed the extraordinary fit between animals
and their natural environment, the characteristic Umwelf (or world view)
of each species, and the particular stimuli (or releasers) that catalyze dra-
matic developmental milestones during "critical" or "sensitive" periods.
Ethology has remained to some extent a European rather than an American
specialty. Still, the willingness to sample wider swaths of behavior in
naturally occurring settings had a liberating influence on the types of
concept and the modes of exploration that came to be tolerated in cognitive
studies.
31
I I THE CoGNITIVE REVOLUTION
The seeds planted in the 1950s sprouted swiftly in the next decade. Gov-
ernmental and private sources provided significant financial support. Set-
ting the intellectual tone were the leading researchers who had launched
the key lines of study of the 1950s, as well as a set of gifted students who
were drawn to the cognitive fields, much in the way that physics and
biology had lured the keenest minds of earlier generations. Two principal
figures in this Selling of cognition" were Jerome Bruner and George
11
Miller, who in 1960 founded at Harvard the Center for Cognitive Studies.
The Center, as story has it, began when these two psychologists ap-
proached the dean of the faculty, McGeorge Bundy, and asked him to help
create a research center devoted to the nature of knowledge. Bundy report-
edly responded, "And how does that differ from what Harvard University
does?" (quoted in Bruner 1983, p. 123). Bundy gave his approval, and
Bruner and Miller succeeded in getting funds from the Carnegie Corpora-
tion, whose president at that time, the psychologist John Gardner, was
sympathetic to new initiatives in the behavioral sciences.
Thereafter, for over ten years, the Harvard Center served as a locale
where visiting scholars were invited for a sabbatical, and where graduate
and postdoctorate students flocked in order to sample the newest thinking
in the cognitive areas. A list of visitors to the Center reads like a Who's
Who in Cognitive Science: nearly everyone visited at one time or another,
and many spent a semester or a year in residence. And while the actual
projects and products of the Center were probably not indispensable for
the life of the field, there is hardly a younger person in the field who was
not influenced by the Center's presence, by the ideas that were bandied
about there, and by the way in which they were implemented in subse-
quent research. Indeed, psychologists Michael Posner and Gordon Shul-
man (1979) locate the inception of the cognitive sciences at the Harvard
Center.
During the 1960s, books and other publications made available the
ideas from the Center and from other research sites. George Miller-
together with his colleagues Karl Pribram, a neuroscientist, and Eugene
Galanter, a mathematically oriented psychologist-opened the decade
with a book that had a tremendous impact on psychology and allied fields
-a slim volume entitled Plans and the Structure of Behavior {1960). In it the
authors sounded the death knell for standard behaviorism with its discred-
ited reflex arc and, instead, called for a cybernetic approach to behavior in
terms of actions, feedback loops, and readjustments of action in the light
32
Cognitive Science: The Rrst Decades
of feedback. To replace the reflex arc, they proposed a unit of activity
called a "TOTE unit" (for "Test-Operate-Test-Exit"): an important prop-
erty of a TOTE unit was that it could itself be embedded within the
hierarchical structure of an encompassing TOTE unit. As a vehicle for
conceptualizing such TOTE units, the authors selected the computer with
its programs. If a computer could have a goal (or a set of goals), a means
for carrying out the goal, a means for verifying that the goal has been
carried out, and then the option of either progressing to a new goal or
terminating behavior, models of human beings deserved no less. The com-
puter made it legitimate in theory to describe human beings in terms of
plans (hierarchically organized processes), images (the total available
knowledge of the world), goals, and other mentalistic conceptions; and by
their ringing endorsement, these three leading scientists now made it legit-
imate in practice to abandon constricted talk of stimulus and response in
favor of more open-ended, interactive, and purposeful models.
The impact of this way of thinking became evident a few years later
when textbooks in cognitive psychology began to appear. By far the most
influential was Cognitive Psychology by the computer-literate experimental
psychologist Ulric Neisser (1967). Neisser put forth a highly "construc-
tive" view of human activity. On his account, all cognition, from the first
moment of perception onward, involves inventive analytic and synthesiz-
ing processes. He paid tribute to computer scientists for countenancing talk
of an "executive" and to information scientists for discussing accession,
processing, and transformation of data. But at the same time, Neisser
resisted uncritical acceptance of the computer-information form of analy-
sis. In his view, objective calculation of how many bits of information can
be processed is not relevant to psychology, because human beings are
selective in their attention as a pure channel such as a telephone cannot
be. Neisser expressed similar skeptical reservations about the claims sur-
rounding computer programs:
33
I I THE CoGNITIVE REvOLUTION
it, both the computer and the human mind should be thought of as ~~sym
bol systems"-physical entities that process, transform, elaborate, and, in
other ways, manipulate symbols of various sorts. And, in 1972, Allen
Newell and Herbert Simon published their magnum opus, the monumental
Human Problem Solving, in which they described the ugeneral problem solver"
programs, provided an explanation of their approach to cognitive studies,
and included a historical addendum detailing their claims to primacy in
this area of study.
Textbooks and books of readings were appearing in other subfields of
cognitive science as well. An extremely influential collection was Jerry
Fodor and Jerrold Katz's collection, The Structure of Language (1964), which
anthologized articles representing the Chomskian point of view in philoso-
phy, psychology, and linguistics, and attempted to document why this
approach, rather than earlier forays into language, was likely to be the
appropriate scientific stance. In artificial intelligence, Edward Feigenbaum
and Julian Feldman put out a collection called Computers and Thought (1963),
which presented many of the best-running programs of the era; while their
collection had a definite ~~carnegie slant," a rival anthology, Semantic Infor-
mation Processing, edited by Marvin Minsky in 1968, emphasized the M.I.T.
position. And, in the area of cognitive anthropology, in addition to influen-
tial writings by Kimball Romney and Roy D'Andrade (1964), Stephen
Tyler's textbook Cognitive Anthropology made its debut in 1969.
But by 1969, the number of slots in short-term memory had been
exceeded-without the benefit of chunking, one could no longer enumer-
ate the important monographs, papers, and personalities in the cognitive
sciences. (In fact, though my list of citations may seem distressingly long,
I have really only scratched the surface of cognitive science, circa 1970.)
There was tremendous activity in several fields, and a feeling of definite
progress as well. As one enthusiastic participant at a conference declared:
When the amount of activity in a field has risen to this point, with
an aura of excitement about impending breakthroughs, human beings
34
CogniHve Science: The First Decades
often found some sort of an organization or otherwise mark the new
enterprise. Such was happening in cognitive science in the early and mid-
dle 1970s. The moment was ripe for the coalescing of individuals, interests,
and disciplines into an organizational structure.
At this time, fate intervened in the guise of a large New York-based private
foundation interested in science-the Alfred P. Sloan Foundation. The
Sloan Foundation funds what it terms "Particular Programs," in which it
invests a sizable amount of money in an area over a few years' time, in the
hope of stimulating significant progress. ln the early 1970s, a Particular
Program had been launched in the neurosciences: a collection of disciplines
that explore the nervous system-ranging from neuropsychology and
neurophysiology to neuroanatomy and neurochemistry. Researchers
drawn from disparate fields were stimulated by such funding to explore
common concepts and common organizational frameworks. Now Sloan
was casting about for an analogous field, preferably in the sciences, in
which to invest a comparable sum.
From conversations with officers of the Sloan Foundation, and from
the pubiished record, it is possible to reconstruct the principal events that
led to the Sloan Foundation's involvement with cognitive science. In early
1975, the foundation was contemplating the support of programs in several
fields; but by late 1975, a Particular Program in the cognitive sciences was
the major one under active consideration. During the following year, meet-
ings were held where major cognitive scientists shared their views. Possi-
bly sensing the imminent infusion of money into the field, nearly every
scientist invited by the Sloan Foundation managed to juggle his or her
schedule to attend the meetings. Though there was certainly criticism
voiced of the new cognitive science movement, most participants (who
were admittedly interested parties) stressed the promise of the field and the
need for flexible research and training support.
While recognizing that cognitive science was not as mature as
neuroscience at the time of the foundation's commitment to the latter
field, officers concluded that "nonetheless, there is every indication,
confirmed by the many authorities involved in primary explorations, that
many areas of the cognitive sciences are converging, and, moreover, there
is a correspondingly important need to develop lines of communication
35
I I THE CoGNITIVE REVOLUTION
from area to area so that research tools and techniques can be shared in
building a body of theoretical knowledge" (Sloan Foundation 1976, p. 6).
After deliberating, the foundation decided to embark on a five-to-seven-
year program, involving commitments of up to fifteen million dollars.
(This commitment was ultimately increased to twenty million dollars.)
The investment took the form, initially, of small grants to many research
institutions and, ultimately, of a few large-scale grants to major universi-
ties.
Like the spur provided by the Macy Foundation a generation earlier,
the Sloan Foundation's initiative had a catalytic effect on the field. As more
than one person quipped, "Suddenly I woke up and discovered that I had
been a cognitive scientist all of my life." In short order the journal Cognitive
Science was founded-its first issue appearing in January 1977; and soon
thereafter, in 1979, a society of the same name was founded. Donald
Norman of the University of California in San Diego was instrumental in
both endeavors. The society held its first annual meeting, amid great fan-
fare, in La Jolla, California, in August 1979. Programs, courses, newsletters,
and allied scholarly paraphernalia arose around the country and abroad.
There were even books about the cognitive sciences, including a popular
account, The Universe Within, by Morton Hunt (1982) and my own historical
essay, also supported by the Sloan Foundation.
Declaring the birth of a field had a bracing effect on those who discov-
ered that they were in it, either centrally or peripherally, but by no means
ensured any consensus, let alone appreciable scientific progress. Patrons
are almost always necessary, though they do not necessarily suffice, to
found a field or create a consensus. Indeed, tensions about what the field
is, who understands it, who threatens it, and in what direction it ought to
go were encountered at every phase of the Sloan Foundation's involvement
(and have continued to be to this day).
Symptomatic of the controversy engendered by the Sloan Founda-
tion's support of research in cognitive science was the reaction to a report
commissioned by the foundation in 1978. This State of the Art Report
(soon dubbed "SOAP" for short) was drafted by a dozen leading scholars
in the field, with input from another score of advisers. In the view of the
authors, "What has brought the field into existence is a common research
objective: to discover the representational and computational capacities of
the mind and their structural and functional representation in the brain"
(1978, p. 6). The authors prepared a sketch of the interrelations among the
six constituent fields-the cognitive hexagon, as it was labeled. Through
the use of unbroken and broken lines, an effort was made to indicate the
connections between fields which had already been forged, and to suggest
the kinds of connection which could be but had not yet been effected.
36
Cognitive Science: The First Decades
I
I
I
I
Artificial
Intelligence
37
I I THE CoGNITIVE REVOLUTION
emerge, thanks to the largesse of the Sloan Foundation, or for some latter-
day Newton or Darwin to bring order into the field of cognitive science.
In the absence, however, of either of these miraculous events, it is left to
those of us who wish to understand cognitive science to come up with our
own tentative formulation of the field. In the opening chapter of this book,
I presented a working definition of cognitive science and alluded to five
key components of the field. Now that I have sketched out some of the
intellectual forces that led to the launching of cognitive science some three
decades ago, I want to revisit these themes in somewhat more detail, in
order to consider some of their implications as well as some of their
problematic aspects. I will then conclude this introductory part by describ-
ing the paradox and the challenge standing at the center of contemporary
cognitive science.
Representations
Cognitive science is predicated on the belief that it is legitimate-in
fact, necessary-to posit a separate level of analysis which can be called
the "level of representation." When working at this level, a scientist traffics
in such representational entities as symbols, rules, images-the stuff of
representation which is found between input and output-and in addition,
explores the ways in which these representational entities are joined,
transformed, or contrasted with one another. This level is necessary in
order to explain the variety of human behavior, action, and thought.
In opting for a representational level, the cognitive scientist is claiming
that certain traditional ways of accounting for human thought are inade-
38
Cognitive Science: The Rrst Decades
quate. The neuroscientist may choose to talk in terms of nerve cells, the
historian or anthropologist in terms of cultural influences, the ordinary
person or the writer of fiction in terms of the experiential or phenomeno-
logical level. While not questioning the utility of these levels for various
purposes, the cognitive scientist rests his discipline on the assumption that,
for scientific purposes, human cognitive activity must be described in
terms of symbols, schemas, images, ideas, and other forms of mental repre-
sentation.
In terms of ordinary language, it seems unremarkable to talk of human
beings as having ideas, as forming images, as manipulating symbols, im-
ages, or languages in the mind. However, there is a huge gap between the
use of such concepts in ordinary language and their elevation to the level
of acceptable scientific constructs. Cautious theorists want to avoid posit-
ing elements or levels of explanation except when absolutely necessary;
and they also want to be able to describe the structure and the mechanisms
employed at a level before "going public" with its existence. While talk
about the structure and mechanisms of the nervous system is relatively
unproblematic-since its constituent units can (at least in principle) be
seen and probed-agreement to talk of structure and processes at the level
of mental representation has proved far more problematic.
Critics of the representational view are generally drawn from behavi-
orist ranks. Wielders of Ockham's razor, they believe that the construct of
mind does more harm than good; that it makes more sense to talk about
neur_ological structures or about overt behaviors, than about ideas, con-
cepts, or rules; and that dwelling on a representational level is unnecessary,
misleading, or incoherent.
Another line of criticism, less extreme but ultimately as crippling,
accepts the need for common-sense talk about plans, intentions, beliefs,
and the like but sees no need for a separate scientific language and level
of analysis concerned with their mental representation: on this point of
view, one should be able to go directly from plans to the nervous system,
because it is there, ultimately, that all plans or intentions must be repre-
sented. Put in a formula, ordinary language plus neurology eliminate the
need for talk of mental representations.
Of course, among scholars who accept the need for a level of represen-
tation, debates still rage. Indeed, contemporary theoretical talk among
"card-carrying" cognitive scientists amounts, in a sense, to a discussion of
the best ways of conceptualizing mental representations. Some investiga-
tors favor the view that there is but a single form of mental representation
(usually, one that features propositions or statements); some believe in at
least two forms of mental representation-one more like a picture (or
image), the other closer to propositions; still others believe that it is possi-
39
I I THE CoGNITIVE REVOLUTION
Computers
While not all cognitive scientists make the computer central to their
daily work, nearly all have been strongly influenced by it. The computer
serves, in the first place, as an "existence-proof": if a man-made machine
can be said to reason, have goals, revise its behavior, transform informa-
tion, and the like, human beings certainly deserve to be characterized in
the same way. There is little doubt that the invention of computers in the
1930s and 1940s, and demonstrations of "thinking" in the computer in the
1950s, were powerfully liberating to scholars concerned with explaining
the human mind.
In addition to serving as a model of human thought, the computer also
serves as a valuable tool to cognitive scientific work: most cognitive scien-
tists use it to analyze their data, and an increasing number attempt to
simulate cognitive processes on it. Indeed, artificial intelligence, the science
built around computer simulation, is considered by many the central disci-
pline in cognitive science and the one most likely to crowd out, or render
superfluous, other older fields of study.
In principle, it is possible to be a cognitive scientist without loving the
computer; but in practice, skepticism about computers generally leads to
skepticism about cognitive science. To some critics, computers are just the
latest of a long series of inadequate models of human cognition (remember
the switchboard, the hydraulic pump, or the hologram) and there is no
reason to think that today's "buzz-model" will meet a happier fate. View-
ing active organisms as "information-processing systems" seems a radical
mistake to such critics. Computers are seen by others as mere playthings
40
CogniHve Science: The Rrst Decades
which interfere with, rather than speed up, efforts to understand human
thought. The fact that one can simulate any behavior in numerous ways
may actually impede the search for the correct description of human behav-
ior and thought. The excessive claims made by proponents of artificial
intelligence are often quoted maliciously by those with little faith in man-
made machines and programs.
Involvement with computers, and belief in their relevance as a model
of human thought, is pervasive in cognitive science; but again, there are
differences across disciplines. Intrinsic involvement with computers is a
reliable gauge of the extent of a discipline's involvement with cognitive
science. Computers are central in artificial intelligence, and only a few
disgruntled computer scientists question the utility of the computer as a
model for human cognition. In the fields of linguistics and psychology, one
will encounter some reservations about a computational approach; and yet
most practitioners of these disciplines do not bother to pick a feud with
computerphiles.
When it comes to the remaining cognitive sciences, however, the
relationship to the computer becomes increasingly problematic. Many
anthropologists and many neuroscientists, irrespective of whether they
happen to use computers in their own research, have yet to be convinced
that the computer serves as a viable model of those aspects of cognition
in which they are interested. Many neuroscientists feel that the brain
will provide the answer in its own terms, without the need for an inter-
vening computer model; many anthropologists feel that the key to
human thought lies in historical and cultural forces that lie external to
the human head and are difficult to conceptualize in computational
terms. As for philosophers, their attitudes toward computers range from
unabashed enthusiasm to virulent skepticism-which makes them a par-
ticularly interesting and important set of informants in any examination
of cognitive science.
41
I I THE COGNITIVE REVOLUTION
42
Cognitive Science: The First Decades
stand out in many pages of cognitive scientific writing. I do not mean that
these traditional questions have necessarily been phrased in the best way,
or even that they can be answered, but rather that they serve as a logical
point of departure for investigations in cognitive science.
In my discussions with cognitive scientists, however, I have found this
precept to be contentious. Nor is it predictable which scientists, or which
science, will agree with a philosophically based formulation of the new
field. Some cognitive scientists from each discipline readily assent to the
importance-indeed, the inevitability-of a philosophical grounding;
while others find the whole philosophical enterprise of the past irrelevant
to their concerns or even damaging to the cognitive scientific effort. We
may well be dealing here with personal views about the utility of reading
and debating classical authorities rather than with fundamental methodol-
ogical aspects of cognitive science. But whatever the reason, cognitive
scientists are scarcely of a single mind when it comes to the importance of
the Meno, of Descartes's Cogito, or of Kant's Critique.
Precisely because the role of philosophy is controversial in the cogni-
tive sciences, it is useful to explore the earlier history of philosophy. Only
such a survey can prove that cognitive scientists-whether or not they are
fully aware of it-are engaged in tackling those issues first identified by
philosophers many decades or even many centuries ago. Scientists will
differ on whether these questions were properly formulated, on whether
philosophers made any significant progress in answering them, and on
whether philosophers today have any proper role in a scientific enterprise.
Indeed, even philosophers are divided on these issues. Still, it is worth
reviewing their positions on these issues, for philosophers have, since
classical times, taken as their special province the definition of human
knowledge. Moreover, they have also pondered the nature and scope of the
cognitive-scientific enterprise, and their conclusions merit serious exami-
nation.
In my own view, each of these symptoms or features of cognitive
science were already discernible in the discussions of the 1940s and were
widespread by the middle 1950s. A cognitive-science text will not neces-
sarily exhibit or illustrate each of the symptoms, but few texts will be
devoid of most of them. What legitimizes talk of cognitive science is the
fact that these features were not in evidence a half-century ago; and to the
extent that they once again pass from the scene, the era of cognitive science
will be at an end.
Comments on the ultimate fate of cognitive science are most properly
left to the conclusion of this study; but as a kind of guidepost to succeeding
chapters, it may be useful to anticipate my principal conclusions. In my
view, the initial intoxication with cognitive science was based on a shrewd
43
I I THE CoGNITIVE REVOLUTION
44
Cognitive Science: The Rrst Decades
to discover or construct the bridges connecting their discipline to neighbor-
ing areas of study-and, specifically, to neuroscience at the lower bound,
so to speak, and to cultural studies at the upper. How to do this (or whether
it can be done at all) is far from clear at this point: but unless the cognitive
aspects of language or perception or problem solving can be joined to the
neuroscientific and anthropological aspects, we will be left with a disem-
bodied and incomplete discipline. Put differently, no one challenges the
autonomy of biology, chemistry, and physics; but unless a single narrative
can be woven from the components of atomic, molecular, and organic
knowledge, the full nature of organic and inorganic matter will remain
obscure.
All this risks getting ahead of our story, however. We have seen in the
preceding pages how different factors present early in the century came
together to form the bedrock of a new discipline. Ultimately, I want to take
a close look at some of the best work in the discipline, so that I can properly
evaluate its current status and its future prospects. To achieve this over-
view, however, it is necessary to consider how the very framing of ques-
tions within cognitive science grows out of philosophical writings of the
past. By the same token, it is necessary to understand the particular histo-
ries, methods, and problems that have characterized the component cogni-
tive sciences. Ultimately this philosophical and historical background has
determined in large meas\.rre the nature and scope of current interdiscipli-
nary cognitive-scientific efforts. In part II of this book, I shall take a careful
look at the several disciplines whose existence made possible the idea of
cognitive science and whose practitioners will determine the success of this
enterprise.
45
PART II
THE COGNITIVE
SCIENCES:
A HISTORICAL
PERSPECTIVE
4
Reason, Experience, and
the Status of Philosophy
Philosophy always buries its undertakers.
-ETIENNE GILSON
Descartes's Mind
49
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
50
Reason, Experience, and the Status of Philosophy
there may perhaps be no such figure anywhere in the world outside of my thought,
nor ever have been, nevertheless the figure cannot help having a certain determi-
nate nature . . . or essence, which is immutable and eternal, which I have not
invented and which does not in any way depend upon my mind. (1951, * p. 61)
What must be the fabric of the nerves and muscles of the human body in order
that the animal spirits therein contained should have the power to move the
members ... what changes are necessary in the brain to cause wakefulness, sleep
and dream; how light, sounds, smells, tastes, and all other qualities pertaining to
external objects are able to imprint on it various ideas by the intervention of the
senses. (Quoted in Wilson 1969, p. 137}
51
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
could now consciously perceive the image and play upon the brain, some-
what as a musician plays upon an instrument (Fancher 1979, pp. 133-34).
Like many who have reflected on the sources of knowledge, Descartes
had once thought that all experiences and thoughts arise through the
senses. But through his meditations, he had come to devalue the senses and
to attribute all thought and creativity to the mind. While he had to admit
that there were sources for the experiences of his senses, he minimized
their significance:
Even as Plato had placed his faith in a mind that could possess (or remem-
ber) all manner of things, Descartes determined that the mind, an active
reasoning entity, was the ultimate arbiter of truth. And he ultimately
attributed ideas to innate rather than to experiential causes:
52
Reason, Experience, and the Status of Philosophy
have been living upon the accumulated capital of ideas provided for them
by the genius of the seventeenth century" (1925, p. 42).
The agenda initially posed by the Greeks, and carried forward during
the seventeenth century by Descartes, came to be fervently debated among
the group of philosophers who earned Whitehead's admiration. The initial
empirical responses by Locke, Hume, and Berkeley, the bold synthesis put
forth by Kant, and the eventual challenge to the Kantian tradition were
principal milestones in the philosophical debates about knowledge. Then,
at the very time when the epistemological tradition associated with these
earlier writers was undergoing the most severe criticism, the computer
came on the scene. Very soon newly discovered philosophical perspectives,
which go by such names as "functionalism," and "intentionality," raised
once again the possibility of a respectable epistemology: a cooperative
scientific effort in which the representational nature of thought could be
countenanced, and where traditional philosophical concerns merged with
the work of cognitively oriented scientists.
Two important themes have recurred in philosophy over the past few
centuries. The first involves the tension between rationalists and empiri-
cists. Those of a rationalist persuasion believe that the mind exhibits
powers of reasoning which it imposes upon the world of sensory experi-
ence; empiricists believe that mental processes either reflect, or are con-
structed on the basis of, external sensory impressions. Both Plato and
Descartes embraced the rationalist pole, while many succeeding empiri-
cists reacted to it. In our own era, behaviorists have clung to empiricism,
while cognitivists are likely to embrace some form of rationalism or a
rationalist-empiricist mix.
A second, discipline-oriented theme concerns the actual status of
philosophy within the world of scholarly disciplines and, in particular, its
relation to science. Once again, Descartes and Plato share a common per-
spective: from their confident stance, philosophical reflection is the pri-
mary pursuit, while the observations of more empirically oriented scholars
are given less credence. (Although science was only beginning to be estab-
lished in Descartes's time, he carried out scientific investigations and surely
thought of himself as a scientific worker.) In succeeding generations, the
findings and laws of science became increasingly visible: indeed, some
issues (for example, the essential nature of matter) were so adequately
resolved by the physical sciences that they dropped outside the purview
of philosophy. Eventually many, if not most, philosophers felt the need
both to keep up with scientific discoveries and to justify their activities in
a scientifically respectable manner. In recent decades, as the dominance of
science continues to grow, the value of philosophical investigations has
again been challenged.
From the perspective of some cognitivists, the rise of empirically
53
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
Let us then suppose the mind to be, as we say, white Paper, void of all
Characters, without any Ideas: How comes it to be furnished? Whence comes it by
that vast store, which the busy and boundless Fancy of Man has painted on it, with
an almost endless variety? Whence has it all the materials of Reason and Knowl-
edge? To this I answer, in one word, From Experience. (Quoted in Herrnstein and
Boring 1965, p. 584)
54
Reason, Experience, and the Status o/ Philosophy
These words sum up the empiricist worldview.
While (on the empiricist account) the experience of the world begins
with perception, it assuredly does not end there. Locke devoted considera-
ble effort to distinguishing among various external qualities (which he
called "primary" and "secondary," "simple" and "complex"). He stressed
how complex ideas grow out of simple ones and the various ways in which
ideas come to be associated with one another. He described the process by
which words come to stand for ideas and make possible abstract or general
ideas-for example, the notion of the general idea of a triangle (as opposed
to a specific triangle with its unique sides and angles). And, in the end, he
posited a person or self able to appreciate these ideas, "a thinking intelligent
being, that has reason and reflection and can consider itself as itself, the
same thinking thing in different time and places" (Copleston 1964, p. 111 ).
Locke's epistemology focused on the opposite end of the universe
from Descartes's-on the experience of objects in the external world. Yet,
Locke was able in the end to construct an organism capable of abstraction
and generalization, one deriving knowledge of a certain kind from the
interplay of these ideas (rather than from a simple comparison with experi-
ence) and culminating in a rational, conscious self. It is not fanciful to
suggest that the Lockean individual bears a strong family resemblance to
the Cartesian individual. But they differ profoundly in the methods by
which each has been achieved.
55
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTivE
thus undermined confidence in a rational order, in scientific explanations
of causal appearances.
He was equally brutal toward the Cartesian belief in the centrality of
the mind, questioning the existence of a substance or locale called "the
mind," in which perceptions inhere. To Hume, we know the mind only as
we know matter-by perception. In fact, the mind is best thought of as an
abstract name for a series of ideas-as a kind of theater where several
perceptions successively make their appearance: there is no observing or
controlling soul behind the processes of thought (Wolff 1967, p. 92). As
the historian Will Durant once quipped (1926), Hume destroyed the mind
as quickly as Berkeley destroyed matter.
Despite the rampant skeptical strain among these latter British empiri-
cists, they produced a set of interlocking issues-the nature of sensory
experience, the classification of objects, the role of language, the status of
the individual conscious self-which budding philosophers were expected
to ponder (issues also that are presented to today's fledgling cognitive
scientists).
Before long, while individual positions continued to change, the belief
that these were the central issues was assumed by all participants in the
debate. In fact, the belief that our conceptual apparatus is built up from
qualities perceived in the external world became so entrenched that for
many years researchers, no less than laypersons, had difficulty in coming
up with rival accounts of why we perceive the world in the way we do.
But, with all that, the nature of mind continued to dominate discussion.
Even Hume, the most skeptical of all, adhered to this central agenda:
By the late eighteenth century, the German scholar Immanuel Kant was
faced with rival alternatives: one, favored by the British empiricists,
viewed thought as merely an instrument to reflect or build upon mundane
56
Reason, Experience, and the Status of Philosophy
experience; the other, favored by Descartes, Leibnitz, and others in the
Continental tradition, stressed the universal province of thought as the
organizer and revealer of all possibilities. Indeed, the German Gottfried
Leibnitz had directly answered the empiricists: addressing Locke's state-
ment that there is nothing in the mind that has not been in the senses,
Leibnitz added the telling Cartesian rejoinder, "Nothing but the mind
[intellect] itself." The empiricists were suspicious of a priori statements and
proofs (which they dismissed as unrevealing tautologies), while the ration-
alists searched for universal principles embodied in pure thought.
In his monumental Critique of Pure Reason, first published in 1781, Kant
strove to synthesize these rationalist and empiricist points of view. To
accomplish this task, he had to confront the question whether there might
exist knowledge that is necessarily so-hence, a priori- but is also in some
sense dependent upon experience, not just conjured up t~utologically in
the mind. He had to apply the systematic analysis of the Leibnitz and
Cartesian tradition to the vagaries of daily experience, which had seemed
regnant to Locke, Hume, and Berkeley. Kant chose to examine both ends
of this polarity: to understand the nature of experience and, even more
important, to dissect the nature of mind.
In what was probably the crucial step in this process, Kant had to
understand what permits the mind to apprehend experience in the way it
does, and to yield necessary knowledge. In analyzing what he called the
synthetic a priori, Kant had to show how knowledge begins with experience
(thus not being purely analytic) and yet does not arise out of, or come from,
it (thus not being purely a posteriori). He had to explain the sources of
arithmetic, geometry, Newtonian physics, traditional logic-those ulti-
mate achievements of human minds, which seemed beyond dispute-that
is, necessary-once they had been discovered.
A point of departure for Kant was the individual ego-the individual
with his own awareness and judgment. In this sense, Kant was a Cartesian:
he began with the knowledge of the ego-the transcendental self. "I think"
must be capable of accompanying all propositions. Nor was this transcend-
ing self a passive instrument: Kant, more than did his predecessors, saw
the mind as an active organ of understanding which molds and coordinates
sensations and ideas, transforming the chaotic multiplicity of experience
into the ordered unity of thought. The transcendental self is always the
active subject: unknown in itself, following rules built into its own opera-
tion, it makes our experience possible.
Nonetheless, this self, this organizing mental entity, does not operate
in autistic fashion. It depends upon and is stimulated by the outside world.
The closest we can get to truth are the assertions we can make about
information that arises under optimal conditions through our sensible
natures. On these assumptions Kant developed a framework detailing how
57
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
knowledge proceeds from and builds out of the "raw" materials of sensory
experience.
Kant's model can be parsed into several parts. Initially, there is the
concrete sensory world-the sensory manifold-which exists external to
the subject and from which one's knowledge must begin. But this sensory
world cannot be perceived directly: there is no privileged access to the
thing itself (das Ding an sich). We must deal always with phenomena-
appearances-and not with noumena-the unknowable external world.
Any phenomenon consists of sensations, which are caused by particular
objects themselves: the form of the phenomenon is due to our subjective
apparatus which orders the manifold in certain ways. The actual forms of
intuition which we use as human beings in apprehending the world are our
"spectacles." For instance, we must see things in terms of their embedded-
ness in space and in time: we have no choice.
But over and above the immanent properties of space and time, our
understanding brings to bear a set of what Kant (following Aristotle) called
"categories of thought." These elementary concepts of the pure under-
standing-such as quantify (unity, plurality, and totality); qualify (reality,
negation, and limitation); relation (substance and accident, cause-and-
effect, and reciprocity); modality (possibility, existence, and necessity)-
constitute the mental equipment, the pure synthesizing concepts with
which human understanding is endowed. These alone allow the individual
to make sense of his experiences.
These categories seem remote from bright red patches or fresh red
cherries-examples that have dominated empiricists' accounts of sensory
experience. Kant may have thought so as well, because he devised yet
another level of analysis-that of the schemas, or schemata-interposed be-
tween the raw sensory information and the abstract a priori categories. As
he put it, with characteristic obscurity, "This representation of a universal
procedure of the imagination in providing an image for a concept, I entitle
the schema of this concept" (quoted in Wolff 1967, p. 76). In devising this
explanatory apparatus, Kant sought to determine the circumstances under
which the categories can find concrete employment. A schema serves as a
mediating representation which is intellectual in one sense, sensible in
another. Thus, a schema is directly activated in terms of sensory experience
and yet can be plausibly thought to provide an interpretation of that
experience. As a cognitive scientist might put it today, Kant had entered
the world of "mental representation."
While the nature and the operation of the schemas pose difficulties
even for Kantian scholars, epistemologists need such a level of analysis to
deal with particular instances while mobilizing the abstract categories. The
schemas are in part rules and thereby are linked to pure understanding; but
58
Reason, Experience, and the Status of Philosophy
they are also in part images and so are linked to empirical perception. The
schema of each category determines the condition by which it is applicable
to the objects of experience in general. Thus, the schema for the category
of quantity is number; the schema for quality is degree of intensity; the
schema for relation is permanence in time; and so on. Moreover, the
schemas are also found at the level of concrete experience. Because we can
apply the concept of dog to Fido, we must be able to produce in our
imagination the schematic representation of a dog-the schema of the
concept here being distinguishable from the concept itself. Schematic the-
ory thus demonstrates how the categories can have empirical conse-
quences.
While Kant did not publish empirical results of any kind, and his
writing on these topics is notorious for its difficulty, his thinking has left
its mark on most theoretical writings in cognitive science today. What
Kant was groping for (in my view) was a way of describing the level of
representation, for a terminology that might account for the way in which
knowledge must be represented in any entity so as to relate to the physical
world, on the one hand, and to the world of the inborn mental architecture,
on the other. And while we may use less tortuous language nowadays for
addressing these issues, I think that no one has gone beyond Kant in
sketching the nature of the problem and in proposing a plausible solution,
on the conceptual level, for how knowledge is possible-or, as he would
have put it, for the conditions for the possibility of knowledge.
Kant had no illusions about the enormousness of the task-but he had
a good deal of confidence about his capacity to negotiate it. He declared
in the preface to the first edition of the Critique of Pure Reason, "I venture to
assert that there is not a single metaphysical problem which has not been
solved, or for the solution of which the key at least has not been supplied"
(quoted in Russell 1961, p. 680). In the preface to the second edition, he
compared himself to Copernicus and opined that, by placing knowledge
in the mind of the active thinker, he had effected a Copernican revolution
in philosophy.
Kant's belief in his own powers eventually proved contagious, and
many philosophically oriented scholars came to believe that he had deli-
mited the domain within which the acquisition of knowledge is possible.
Kant's charting of territory included three reasons why a science of psy-
chology was not possible: the fact that the mind is affected while studying
itself; the lack of spatial extent essential for any experimentation; and the
absence of a mathematical basis, necessary for all science. Kant concluded
that "psychology can, therefore, never become anything more than a his-
torical (and, as such, as much as possible) systematic natural doctrine of
the internal sense, i.e., a natural description of the soul, but not a science
59
II I THE CoGNITIVE SciENCEs : HISTORICAL PERSPECTIVE
Kant's program, his conclusions, and even his warnings exerted an unpar-
alleled effect on ensuing research. Whereas each of the earlier epistemolo-
gists had soon aroused a formidable opponent, Kant's epistemology re-
mained without major competition for years. He had apparently reconciled
the two competing strands in philosophical writing-the primacy of
thought and the primacy of experience; and his argument was sufficiently
difficult, his critiques of standard rationalist and empiricist positions suffi-
ciently powerful, his examination of the nature of mentation sufficiently
revolutionary, his view of philosophy sufficiently reassuring, that it was
many years before the weaknesses in his formulation became apparent.
Thanks to Kant, philosophy for a time occupied a special place in the
firmament of scholarly disciplines. In light of his supposed demonstration
of what knowledge was, of how its attainment was possible, and of why
Newtonian science was necessary, Kant conferred upon philosophy the
status of a "super" or "meta" discipline. Every scholar was expected to
study this foundational subject, which dealt with the most fundamental
aspects of knowledge; and no educated person would admit to not having
read Kant. Anyone who entertained grave doubts about the utility of
philosophical analysis usually kept silent. And it continued to be assumed
that science and philosophy could march onward together.
It was not the philosophers who adopted Kant's speculative mode, but
rather logicians, who eventually challenged Kantian epistemology. Kant
had declared that, since Aristotle, logic "has not been able to advance a
single step and is to all appearances a closed and completed body of
doctrine" (quoted in Abel1976, p. 52). In fact, however, in the century and
a quarter following Kant, thinkers like George Boole, Gottlob Frege, Gi-
useppe Peano, Charles Sanders Peirce, and, ultimately Bertrand Russell
and Alfred North Whitehead had effected many reformulations in logic.
Indeed, rather than being a set of approximate procedures, logic became
a non-empirical science, into which new findings could be regularly ab-
sorbed. There were equally dramatic advances in mathematics and physics,
as scholars like Niels Bohr, George Cantor, Albert Einstein, and Max
Planck rewrote our understanding both of the world of mathematics and
60
Reason, Experience, and the Status of Philosophy
of the physical universe. Eventually, progress in the fields that Kant
held in highest regard called into question many of his fundamental as-
sumptions: the given nature of time and space, the source of mathematical
propositions, the unmodifiable rules of logic, the impossibility of
psychology.
The work that had the greatest effect on the mainstream of epistemol-
ogy was carried out in Cambridge, England, at the beginning of the twen-
tieth century by Alfred North Whitehead and Bertrand Russell. These
mathematically oriented logicians had been impressed by the work of the
German logician Gottlob Frege who had undertaken to show that arithme-
tic could be reduced to logic. As Frege had written, "Even an inference like
that from n to n + 1, which on the face of it is peculiar to mathematics,
is based on the general laws of logic" (quoted in Hanfling 1972, p. 103).
Building on Frege's insights, Whitehead and Russell sought to derive all
of mathematics from the basic laws of logic, as these laws had been refor-
mulated in the century since Kant's time. Indeed, this program was in part
explicitly anti-Kantian, for the philosophers wanted to discredit the syn-
thetic a priori notion of mathematical knowledge as being dependent upon
experience. With respect to this part of their program, they were largely
successful.
But Whitehead and Russell's overall program exerted its most revolu-
tionary effect on philosophy. Their work led to much closer ties among
empirical science, logic, and mathematics, so that the boundaries between
these pursuits could no longer be sharply delineated. Whereas Kant saw
philosophy as the ultimate foundational discipline, the standard by which
other sciences had to be judged, Whitehead and Russell saw all of these
disciplinary forays as part of the same larger enterprise. Indeed, Russell
believed that many-if not most-traditional philosophical questions
could be expressed in logical terms and either solved in these terms, or
shown to be insoluble. As he confidently declared:
Modern analytical empiricism ... differs from that of Locke, Berkeley, and
Hume by its incorporation of mathematics and its development of a powerful
logical technique. It is thus able, in regard to certain problems, to achieve definite
answers, which have the quality of science rather than of philosophy. It has the
advantage, as compared with the philosophies of the system-builders, of being able
to tackle its problems one at a time, instead of having to invent at one stroke a block
theory of the whole universe. Its methods, in this respect, resemble those of
science. I have no doubt that, in so far as philosophical knowledge is possible, it
is by such methods that it must be sought: I have also no doubt that, by these
methods, many ancient problems are completely soluble .... Take such questions
as: What is number? What are space and time? What is mind, and what is matter?
I do not say that we can here and now give definitive answers to all these ancient
questions, but I do say that a method has been discovered by which, as in science,
61
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
we can make successive approximations to the truth, in which each new stage
results from an improvement, not a rejection, of what has gone before. (Russell
1961, pp. 788, 789)
62
Reason, Experience, and the Status of Philosophy
tions about the world (Kenny 1973). At the same time that it clarifies the
propositions of natural science, philosophy can also expose as nonsense
metaphysical questions about existence.
As is well known, Wittgenstein himself came first to doubt and then
to renounce virtually completely his "picture theory" of language and the
world (Wittgenstein 1968). But his program of analyzing language logi-
cally and trying to relate that analysis to the perceived world became a
pivotal ingredient of the program of philosophy adopted by the "Vienna
circle" of logical empiricists in the period between the world wars. Scholars
like Herbert Feigl, Otto Neurath, Morris Schlick, and especially Rudolf
Camap tried in earnest to tie together what A.]. Ayer called "language,
truth, and logic" (1936). The general goal of the circle was to see which
traditional philosophical questions could be rephrased in formal terms and
which could not be. Those that could not be (for instance, issues about
celestial angels) were branded as metaphysical and banned forthwith from
further discussion. Those propositions that lent themselves to treatment in
logical terms (for instance, ones about geometric angles) would then be
examined to see whether they could be verified and thus added to our
storehouse of what is true.
One major ingredient of the program of the logical empiricists was
verificafionism, a doctrine that assumed that empirical (nonlogical) state-
ments can be verified under ideal conditions of inquiry. All that is needed
is a way of measuring and verifying what is being talked about. As Carnap
put it, one must observe the circumstance under which people use a propo-
sition in order to determine the truth of that proposition. Indeed, the
meaning of a proposition is its method of verification. Another doctrine of
the program was physicalism. The Vienna circle believed that propositions
ordinarily construed as referring to mental states turn out to be logically
equivalent to those referring to overt behavior. Indeed, every sentence of
psychology could be reformulated as a description of the physical behavior
of humans or other animals. Finally, the Vienna circle rejected the notion
of philosophy as a special or foundational area of study: as far as these
philosophers were concerned, there was just one empirical science, of
which philosophy was part (Ayer 1982, p. 128); its role was to describe and
criticize the language of science.
Armed with this set of ideas, Rudolf Carnap determined to put into
practice the logical-empiricist beliefs. Russell had talked about defining
the world from experience, through logical construction; and Carnap
sought to translate, into the language of sensory data, all sentences pertain-
ing to the world. For every sentence about material objects, a correspond-
ing sentence about sensory data can express the basic phenomenal ele-
ments of the experience of the object.
Having devised sensory-centered sentences of this sort, the next step
63
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
The Aufbau [the logical construction of the world) brings to philosophy the
powerful techniques of modern logic, along with unprecedented standards of ex-
plicitness, coherence, and rigor. It applies to [philosophy] the new methods and
principles [which had been brought to bear] upon mathematics .... The potential
importance to philosophy is comparable to the importance of the Euclidean de-
ductive method into geometry.... [It] is still one of the fullest examples we
have of the logical treatment of problems in nonmathematical philosophy. (1972,
p. 22)
Thus, with Russell and Camap, Goodman believes that the techniques
from logical analysis have the potential to solve many-even if not all-
existing philosophical problems, and that attacking the problems one at a
time is the best means for achieving success.
There is another reason the saga of logical empiricism is highly rele-
vant to a history of cognitive science. Inspired by the Whitehead-Russell
tradition, Camap and his colleagues sought to express everyday scientific
findings in terms of the basic elements of logic. Loosely speaking, we might
see this effort as a ferreting out of the logical structure of science and, more
especially, of the language used in science-as a study of the syntax of
science. As I see it, a major ingredient in ongoing work in the cognitive
sciences has been cast in the image of logical empiricism: that is, the vision
of a syntax-a set of symbols and the rules for their concatenation-that
might underline the operations of the mind (and a correlative discomfort
with issues of mental content). Thus, when Noam Chomsky (1965) posits
the basic operations of a grammar, when Richard Montague (Thomason
1974) examines the logic of semantics, when Allen Newell and Herbert
Simon (1972) simulate human reasoning on a computer, or when Jerome
Bruner (1973) and George Miller (1956) seek to decipher the rules of
classification, or "chunking," they are trying to decipher a logic-perhaps
64
Reason, Experience, and the Status of Philosophy
the logic-of the mind. This vision comes through even more clearly in the
writings of Jerry Fodor, who explicitly searches for a "language of
thought" and even appropriates certain of Camap's methods. Thus, a
model that proved inadequate for the scientific enterprise as a whole still
motivates research in circumscribed cognitive domains.
In applying logical methods to the world of empirical experience, the
members of the Vienna circle combined the spirit of rationalism and em-
piricism. They also sought to establish a place for philosophy within the
world of laboratory science by providing the tools for analyzing scientific
statements and practices. In hindsight, their procedures were too artifi-
cially constrained, and their idea of theory too tied to behaviorist and
positivist strictures. And yet even those cognitivists who see themselves
as doctrinarily opposed to the letter of the Vienna program often partake
of the spirit of their approach.
During the late 1930s and early 1940s, a group of philosophers met regu-
larly at Harvard. Among them were the Harvard philosophers Nelson
Goodman and W. V. 0. Quine, as well as such frequent visitors as Rudolf
Camap, Bertrand Russell, and Alfred Tarski, all three from war-tom
Europe. These philosophers had all been involved with logical empiricism
but now subjected nearly all of its assumptions to increasingly severe
criticism. First of all, they questioned whether it is possible to talk about
"raw information": that is, pure sensory data which can be inspected and
built upon. Knowledge came to be seen increasingly as a matter of using
propositions, and it was now dubious practice to talk about knowing the
pure, the immediate, the given. Even more tellingly, the whole logical
apparatus devised to analyze "meaning" and "truth" came to be chal-
lenged. While the distinction between analytic and synthetic truths had
been cherished since the time of Kant, Quine {1953) showed that it is
ultimately untenable: the notion of meaning itself is not clear enough to
justify the attribution to certain statements of "true by virtue of meaning
alone" or "true by definition." Moreover, the logical (or analytic) compo-
nents of a scientific theory cannot be sufficiently disentangled from the
empirical components to allow them to be regarded as subject to different
criteria of truth. After all, Euclid's principles would certainly seem to have
been a priori truths; and yet their truth was ultimately undermined by
65
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
Not a single one of the great positive theses of Logical Empiricism (that
Meaning is Method of Verification; that metaphysical propositions are literally
without sense; that Mathematics is True by Convention) has turned out to be
correct. It detracts from the excitement of the fact that, by turning philosophical
theses into linguistic ones [as Carnap had tried to do] ... one can make philosophy
more scientific and settle the truth value of philosophical propositions by hard
scientific research, if the results one obtains are uniformly negative. (1975a, p. 20)
Still, he softened this verdict, in ways revealing for this history, by declar-
ing, "Even if [logical empiricism] failed, modern symbolic logic, a good
deal of modern language theory, and part of contemporary cognitive sci-
ence were all off-shoots of these attempts" (1984, p. 274).
Critics like Quine and Putnam regretted the collapse of logical empiri-
cism and have been sympathetic toward cognitive science, perhaps sensing
that this movement shares at least some of the methods and aspirations of
earlier philosophical work. But by the end of the 1940s, far more strident
voices were being raised about the tradition that had begun with Descartes
and was, after many twists and turns, still discernible in the logical-empiri-
cist camp. The trio of names most vividly associated with this criticism are
Gilbert Ryle, Ludwig Wittgenstein (in his later writings), and]. L. Austin.
Gilbert Ryle
In his justly celebrated The Concept of Mind (1949), Ryle explicitly con-
fronted the "official doctrine" of mentalism that had begun with Descartes.
As Ryle described this doctrine, it entailed the belief that everyone has a
mind and a body; minds are not in space and their operations are not
subject to mechanical laws; the workings of the mind are private and are
accessible only to the person alone; there are, in fact, two different kinds
of existence or status-whatever happens may have a physical or a mental
existence. "Mental happenings occur in insulated fields, known as 'minds'
and there is ... no direct causal connection between what happens in one
mind and what happens in another" (1949, p. 13). Ryle dubbed this doc-
trine, with its many associated assumptions and linguistic turns, as the
"dogma of the Ghost in the Machine." He declared, "I hope to prove that
66
Reason, Experience, and the Status of Philosophy
it is entirely false, and false not in detail but in principle" (p. 16), and
devoted a 330-page book to justifying that hope.
Ryle's basic claim was that talk of mind involves a category mistake.
It is perfectly all right to speak about "minds"; but we must not fall into
the trap of thinking that there is a place called "the mind," with its own
locations, events, and so on, any more than we assume that there is a place
called "the university" apart and separate from buildings, roads, lawns,
persons, and other physically specifiable entities. To talk of mind as if it
has a separate existence, was, to Ryle, mistakenly to treat an entity of one
sort (an abstract characterization of a set of dispositions) as if it were itself
another of those dispositions. "The theoretically interesting category-mis-
takes," Ryle pointed out, "are those made by people who are perfectly
competent to apply concepts, at least in the situations in which they are
familiar, but are still liable in their abstract thinking to allocate those
concepts to logical types to which they do not belong" {1949, p. 17). And
the mind becomes such a theoretically interesting mistake because Des-
cartes posited a substance, parallel to body but apart from it, that controls
and is the scene of our mental life.
Having exposed the flaw in the Cartesian position, Ryle proceeded to
show how one can talk, in ways that do not involve category mistakes or
violate the actual state of affairs, about those entities and experiences that
are generally termed "mental." In general, he took a behaviorist stance:
when we talk about a person as having "traits" or inner volitional capaci-
ties, we are simply indicating that people are disposed to behave in certain
ways and are likely to do so in the presence of the appropriate eliciting
circumstances. Ryle questioned whether there are actually happenings in
the mind to which each person has privileged access. Instead, he insisted,
what we can find out about ourselves is in principle no different from the
sorts of things we can find out from other people through observation and
through questioning. To speak of a person's mind is simply to talk of
certain ways in which the incidents of one's life are ordered.
Ryle declared that he was not interested in the issues of how one sees
or understands something if those seeings or understandings involve the
positing of some internal understanding or perceptual mechanisms. The
most that we can do as philosophical analysts is to try to understand the
circumstances under which an individual would report having seen or un-
derstood something; and these circumstances should be accessible across
the full range of reporting individuals. Ryle objected to answers invoking
internal mechanisms ("I understand something because I process certain
information in certain ways") to conceptual questions ("What are the
circumstances under which an individual is likely to report having under-
stood something?"). In his view, the positing of internal mechanisms (a
67
II I THE CoGNITIVE SciENCEs : HisToRICAL PERSPECTIVE
Ludwig Wiffgensfein
Wittgenstein saw that many philosophical conundrums might be
solved by careful attention to the ways in which people use words. In his
earlier writings, he had treated language as a means of understanding the
world-as a privileged means of looking through to the structure of the
world. Now, however, he saw language itself as the spawner of prob-
lematic issues, and the exercise of coming to understand how we use
language as the therapy for philosophical problems. In his later work,
Wittgenstein did not try to solve problems but tried rather to show that
they arise from a network of terms that have evolved in such a way as
to make their disentanglement extremely difficult. As he once com-
mented, his aim in philosophy was to show the fly the way out of the
fly-bottle.
Wittgenstein came to view language as an inherently public or com-
munal activity. One is introduced into language by others in the commu-
nity, and thus one comes to know how to use words. One certainly has
a private experience, for example, of pain, but the use of the word
pain comes from the ways in which, and the circumstances under which,
it is regularly employed by other people. We do not first use the word
pain by naming something we feel; rather, the word is embedded in various
activities that people carry out and in various sorts of things others say in
connection with being hurt.
More generally, Wittgenstein believed, it is instructive to think of
language as a set of games: we proceed from the fact that we are always
involved in many language games-interactions with other individuals in
which we move around sets of linguistic counters; and, like a set of games,
each of these little encounters has its own set of rules. But it is not easy
to ferret out these rules because they overlap with one another: the lan-
guage games constantly mesh. To add to this tangled state of affairs, words
do not have clear and unambiguous meanings. The word game itself has a
family of meanings, with no definition ever sufficient to account for all,
and only all, games. Given the numerous language games occurring at any
one time, and their inherently overlapping nature, it is no wonder that
Wittgenstein despaired of ever solving philosophical problems in the rig-
orous way that he and his Viennese peers had once hoped. It made more
68
Reason, Experience, and the Status of Philosophy
sense to try to dissolve the problems altogether, by showing that they had
been deceptively phrased.
Wittgenstein' s attitude toward traditional problems can be gleaned
from his comments about psychology, particularly the version that he had
encountered during his early studies in Vienna. As he saw it, psychologists
were trying to solve problems they did not understand: rather than being
genuinely scientific, these problems were in fact embedded in certain uses
of language. Wittgenstein illustrated the complex and overlapping manner
in which many mentalistic words-believe, imagine, desire-are actually used.
Instead of trying to explain how each of these putative mental operations
actually "works," it would make more sense for psychologists (a Ia Ryle)
to study the relations among such ways of talking about behavior and
experience. In a pessimistic assessment, Wittgenstein asserted:
f L. Austin
As if one were needed-a final nail was sunk into the coffin of logical
empiricist philosophy by J. L. Austin, another British philosopher inter-
ested in language. Austin demonstrated convincingly that one cannot sim-
ply accept a sentence at face value, as the logical empiricists wanted to do.
Many, if not most, sentences need to be thought of not only in terms of their
literal meaning (ullerance meaning} but also with respect to the use to which
they are put by the deliverer of the utterance (speaker meaning}. As Austin
explained in his William James lectures, How to Do Things with Words:
it has come to be commonly held that many utterances which look like
statements are either not intended at all, or only intended in part, to record or
69
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
impart straightforward information about the facts .... Many traditional philo-
sophical perplexities have arisen through a mistake-the mistake of taking as
straightforward statements of fact utterances which are eilhn (in interesting non-
grammatical ways) nonsensical or else intended as something quite different.
(1962, pp. 2-3)
Freddy [A. J.] Ayer was there, still very keen on the Vienna circle. We all knew
by the time the evening was half-done that Camap and Vienna were finished.
Austin made the distinction between analytic and synthetic propositions a special
case of his more general way of classifying propositions by their illocutionary"
force. It was stunning. (Quoted in Bruner 1982, p. 41)
But I think that at this point it may be more useful to say rather that episte-
mology still goes on, though in a new setting and a clarified status. Epistemology,
or something like it, simply falls into place as a chapter of .psychology and hence
of natural science. It studies a natural phenomenon, viz, a physical human subject.
This human subject is accorded a certain experimentally controlled input-certain
patterns of irradiation in assorted frequencies, for instance-and in the fullness of
• Rlocufion refers to the ends for which a statement is uttered-for example, to give an
answer, to make an announcement, or to make a request.
70
Reaspn, Experience, and the Status of Philosophy
time the subject delivers as output a description of the three-dimensional external
world and its history. The relation between the meager input and the torrential
output is a relation that we are prompted to study for somewhat the same reasons
that always prompted epistemology; namely, in order to see how evidence relates
to theory, and in what ways one's theory of nature transcends any available
evidence. (Quoted in Royce and Rozeboom 1972, p. 18)
Surveying the same scene as Quine, Richard Rorty reaches a far less san-
guine conclusion. In his much-discussed Philosophy and the Mirror of Na-
ture (1979}, Rorty questions the whole enterprise of epistemology. He
suggests a radically different and far more modest view of philosophy, one
that might even undermine the programs pursued by Wittgenstein, Ryle,
and Austin, and certainly one much less ambitious and far less optimistic
than Quine's. Rorty's impressive critique of all of philosophy since the
Greek times not only has intrinsic interest but also raises profound ques-
tions about the enterprise called cognitive science, at least as I have charac-
terized it here.
According to Rorty, philosophers have thought of their discipline as
71
II I THE CoGNITIVE SciENCEs : HISTORICAL PERSPECTIVE
I want to suggest that the concept of mind is the blur with which Western
intellectuals became obsessed when they finally gave up on the blur which was the
theologian's concept of God. The ineffability of the mental serves the same cultural
function as the ineffability of the divine-it vaguely suggests that science does not
have the last word. (Rorty 1982b, p. 31)
72
Reason, Experience, and the Status of Philosophy
pends upon the activity of a quasi-visual faculty-which Rorty dubs the
Mirror of Nature-and consists of an assembly of accurate representations.
These representations are in the mind; and an "inner eye" surveys them
hoping to find some mark that will testify to their fidelity. Though the
empiricists differed with Descartes in many respects, they preserved his
mentalistic notion of the mind as a separate region which inspects ideas.
Locke introduced a fatal confusion when he confounded an explanation of
how information gets into our consciousness-a classical psychological
question-with a justification of why we believe what we believe-a classi-
cal philosophical question.
Unlike his predecessnrs, Immanuel Kant understood the impossibility
of anyone's having direct access to things: that is, one has knowledge of
propositions about objects not about objects themselves. But, in an effort
to locate the most accurate representations, Kant ultimately posited a
special privileged set of representations that cannot be doubted. According
to Rorty's revisionist account, these representations came to be considered
the foundation of all knowledge; and thus Kant granted to philosophers
the pre-eminent position for making statements about the world and for
regulating inquiry.
In Rorty's view, in the years since Kant's time much of philosophy has
sought to retain this vision. But the vision has undergone severe jolts,
thanks to a series of compelling critiques of traditional epistemology.
There was Wittgenstein's attack on the centrality and legitimacy of classi-
cal philosophical problems. There was the pragmatist John Dewey, who
insisted that one should try to use knowledge in a practical way rather than
to strive after the chimera of objective knowledge. There was the
phenomenologist Martin Heidegger, who dissected the various images and
metaphors that have obsessed Western philosophy since Greek times. And
perhaps most tellingly, there were the failures of both the Russell-White-
head program and the Vienna circle to arrive at secure knowledge through
a logical construction from sense data.
Rorty's own interpretation of recent philosophical history relies espe-
cially heavily on criticisms to which I have already alluded. He arrives at
the following conclusion: There is no way to account for the validity of
our beliefs by examining the relation between ideas and their objects:
rather, justification is a social process, an extended conversation, whereby
we try to convince others of what we believe. We understand the nature
of knowledge when we understand that knowledge amounts to justifica-
tion of our belief, and not to an increasingly accurate representation of
reality. As Rorty concludes, "If assertions are justified by society rather
than by the character of the inner representation they express, there is no
point in attempting to isolate privileged representations" (1979, p. 174).
73
II I THE CoGNITIVE SciENCEs : HisTORICAL PERSPECTIVE
Psychologists should be more mechanistic rather than less .... They should
cut straight through the mental to the neurophysiological. (1979, p. 217)
This is to say that if physiology were simpler and more obvious than it is,
nobody would have felt the need for psychology. (p. 237)
We can imagine machines in which it would be easier to find out what the
machine was up to by opening it up and looking than by reading the program.
(p. 238)
74
Reason, Experience, and the Status of Philosophy
If we have psychophysiology to cover causal mechanisms, and the sociology
and history of science to note the occasions on which observation sentences are
invoked or dodged in constructing and dismantling theories, then epistemology has
nothing to do. We would think that this result would be congenial to Quine, but
in fact he resists it. (p. 225)
And, in the context of discussing the effect that his Antipodeans might
have on philosophy, Rorty adds, "The disappearance of psychology as a
discipline distinct from neurology, and similar cultural developments,
might eventually free us from the image of the Mirror of Nature much
more effectively than philosophers' identity theories" (p. 121}.
To summarize, Rorty is unable or unmotivated to come up with any
arguments in principle against psychology, but feels that the discipline
might well never have been invented, and that it may well at some time
disappear, to the regret of few. Neurophysiology seems a much more
secure science on which to base one's hope. The issue of an interdiscipli-
nary cognitive science is not addressed directly by Rorty, but some of his
thoughts on this matter can be gleaned from this telltale aside:
Only the assumption, that one day the various taxonomies put together by,
for example, Chomsky, Piaget, Levi-Strauss, Marx, and Freud will all flow together
and spell out one great Universal Language of Nature-an assumption sometimes
attributed to structuralism-would suggest that cognitive psychology had epis-
temological import. But that suggestion would still be as misguided as the sugges-
tion that, since we may predict everything by knowing enough about matter in
motion, a completed neurophysiology will help us demonstrate Galileo's superior-
ity to his contemporaries. The gap between explaining ourselves and justifying
ourselves is just as great whether a programming language or a hardware language
is used in the explanations. (1979, p. 249)
75
II I THE CoGNITIVE SciENCEs : HisTORICAL PERSPECTivE
76
Reason, Experience, and the Status of Philosophy
Reflecting on this incident some twenty years later, Popper repeated
his claim that there are indeed philosophical problems, and that he may
have even solved some of them. Yet he went on to add, "Nothing seems
less wanted than a simple solution to an age-old philosophical problem.
The view of many philosophers, and, especially, it seems, of Wittgen-
steinians, is that if a problem is soluble, it cannot have been philosophical"
(1974, p. 124).
This confrontation between Popper and Wittgenstein helps to convey
the atmosphere in philosophical circles at the middle of this century. In the
background stood the great figures of the past-Popper's heroes-who
believed that there are genuine philosophical problems, and that if one
ponders them systematically, one should eventually solve them, or at least
make progress toward solving them. Rorty also recognized this group but
believed that these "systematizers" (as he termed them) have been em-
barked on a hopeless and wrong-headed task. Instead, Rorty reserved his
praise for individuals whom he labels "edifiers," who are deeply skeptical
about the legitimacy of such questions and would rather assume a soft,
even teasing approach toward the whole philosophical enterprise. Among
scholars in the edifying camp were those who criticized the logical empiri-
cist approach, including Wittgenstein, Dewey, Austin, and Rorty himself.
As far as I can tell, the systematizers are not about to relinquish their
calling. Thus, a sympathetic reviewer of Rorty's book remarks:
If "epistemology" means the search for such foundation, then its end is in
sight, an end foreseen by Dewey. But if "epistemology" denotes an attempt to
understand the possibility and nature of the various kinds of knowledge and styles
of reasoning, then Plato, Locke, and Dewey are part of a persistent tradition that
has to do with one of the essential characteristics of our civilization. (Hacking 1980,
p. 586)
77
I I I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
Functionalism
One major contributor to the discussion is Hilary Putnam himself. A
mathematically trained philosopher, Putnam has had a long-standing in-
terest in the nature of computers and their implications for thinking. As
he himself recalls, the invention of computing machines was an important
event in the philosophy of mind because it led to the idea of functional
organization (1973, p. 299). This functionalist idea challenges the assertion
that thinking and other "intelligent functions" need to be carried out by
means of the same specified machinery in order to reflect the same kinds
of process. To be sure, before the advent of computing machines, it might
have been tenable to hold that thinking can only occur in human beings,
or in entities with the kind of brain structure that we have. But computers
demonstrated that many of the processes we would once have termed
"thinking" can indeed be realized by mechanisms that are constituted of
entirely different components-transistors or vacuum tubes-rather than
nerves, blood, and tissues.
If there was to be an identity, it obviously could not reside in the
hardware but, as Putnam pointed out, might well occur in the software:
that is, both human beings and machines-and any other form of intelli-
gent life, from anteaters to Antipodeans-could be capable of realizing the
same kinds of program. Thus, the equation occurs at a much higher level
of abstraction-a level that focuses on the goals of a cognitive activity, the
means of processing at one's disposal, the steps that would have to be
taken, the evaluation of the steps, and kindred features.
Indeed, functionalism purported to address one of the most classic
philosophical dilemmas-the mind-body problem. Thought can indeed
occur in a physical apparatus and can be correlated with certain behavior
and yet not have to be identified with the precise class of activities that
78
Reason, Experience, and the Status of Philosophy
happen to be produced (Matthews 1982). It is perfectly legitimate to talk
of mental events, to posit one mental event as causing another, and to do
so without taking a position on whether only brain events have the proper-
ties to define mental states. Of course, in human beings, mental events are
identical to physiological ones, but there is no need to reduce psychological
explanations to the neuro-physiological level. Rather one may counte-
nance a level of explanation that connects psychology to neurology, a
second level that connects it to social factors, and even a third free-
standing level, which is representational.
Intentional Systems
Building on Putnam's pioneering attempt to tackle the issues of
knowledge in an idiom that makes sense in the contemporary cognitive
milieu, certain philosophers in more recent times have put forth their own
analytic schemes. Daniel Dennett begins by distinguishing his own theo-
ries from earlier efforts (Dennett 1978). He does not wish to embrace a kind
of physicalism in which, for every type of mental event, there is a particu-
lar type of physical event in the brain. Nor does he want to adopt a
so-called Turing-machine functionalism, which would imply that all in-
dividuals have the same program. He thus searches for a level of explana-
tion where it is possible to talk about what two individuals or two entities
have in common psychologically without both being the realization of
some single Turing machine.
Enter the intentional system. This concept is designed to play a role
in the legitimation of mental descriptions parallel to the role played by the
abstract notion of a Turing machine in setting down the rules for interpret-
ing an artifact like a computer. Intentionality is seen as the mark of the
mental, and Dennett claims that every mental phenomenon can be de-
scribed in terms of intentional systems.
As Dennett phrases it, one may consider both human beings and a
computer and its programs as agents whose acts one is trying to explain.
Though one can talk about the computer in terms of its physical design
or its actual physical states, it often makes more sense to treat the machine
as if it were significantly like an intelligent human being. In other words,
one attributes rationality and purpose to an intentional system. And so, if,
for example, you are playing a game of chess with a computer, you think
of the program as having at its disposal certain goals, procedures, strate-
gies; and you try to outwit the program just as you would try to outwit
that more conventional kind of intentional system-another person.
The notion of an intentional system, Dennett argues, is relatively
uncluttered and unmetaphysical. It does not touch the issue of the compo-
79
II I THE CoGNITIVE SciENCEs : HisTORICAL PERSPECTIVE
[T]he definition of intentional systems I have given does not say that inten-
tional systems really have beliefs and desires but that one can explain and predict
their behavior by ascribing beliefs and desires to them, and whether one calls what
one ascribes to the computer beliefs or belief-analogues or information complexes
or intentional whatnots makes no difference to the nature of the calculation one
makes on the basis of the ascriptions. (1978, p. 7)
But if this is only a manner of talking, what has been gained by this
apparently daring move? It is likely that Dennett, a student of Ryle' s, would
really rather dispense with this mentalistic line altogether and fall back on a
more trustworthy behaviorist mode of discussion. Still, Dennett concedes
that such discussion really is no longer adequate for scientific purposes. As
he comments in a critical piece on the behaviorist B. F. Skinner, "There is no
reason why intentional terms can't be used provisionally in the effort to
map out the function of the behavioral control system of men and animals
just as long as a way is found to' cash them out' by designing a mechanism to
function as specified" (Dennett 1978, p. 62).
80
Reason, Experience, and the Status of Philosophy
In developing their notions, Putnam and Dennett have exemplified
the contribution of philosophy to cognitive science. Responding to the
various issues raised by the advent of the computer, while drawing on
long-standing discussion about the mind's relation to the body and an
agent's sense of purpose, these authorities have helped to clarify issues
raised in the contemporary science of mind. Still, I think it would be fair
to state that both of these philosophers have entered the terrain of cogni-
tive science with considerable caution. And, as we shall see in later chap-
ters, more vexing issues have been raised, by philosophical critics of artifi-
cial intelligence, such as John Searle and Hubert Dreyfus. For these
reasons, it is salutary to consider the work of Jerry A. Fodor, a full-scale
cognitivist-one philosopher who seemingly has no reservations whatso-
ever about the common fate of philosophers and empirical scientists inter-
ested in issues of mind.
To read Fodor (1975, 198lb) is to feel that the tables of epistemology have
been suddenly and radically turned: the "bad guys" have been propelled
to the position of heroes and some widely heralded common heroes have
been demoted. Following his one-time mentor and present colleague Noam
Chomsky, Fodor finds much to admire in the analysis put forth by Des-
cartes over three hundred years ago. For Fodor, the Cartesian tradition has
the merit that it recognized the existence of mental states and freely per-
mitted mental events to have causal power. Furthermore, it countenanced
the positing of innate ideas-informational content, mechanisms, or prin-
ciples with which the individual is born and which allow the individual
to make sense of experience. As part of his own version of epistemology,
Fodor is vociferously critical of the empiricist tradition. He feels that three
hundred years of empiricist efforts, from Hume to the logical positivists,
have resulted in failure. On Fodor's reading, both empiricists and rational-
ists had accepted that some concepts must be innate, but he sides with the
rationalists in their belief that the human being is born with the knowledge
of many concepts, which, at most, must simply be triggered by the environ-
ment. Fodor also excoriates the Rylean behaviorist position: he, in fact,
wants to claim that the kinds of explanations Ryle rejected are just those
that are needed. In Fodor's view, it is quite a different matter to explain
the circumstances under which an utterance is made from the reasons it
is made; and, in defiance of Ryle, it is part of Fodor's project to invade the
81
I I I THE CoGNITIVE SciENCES : HisTORICAL PERSPECTIVE
mental space: to figure out just why and how we make the statements we
make, given the mental equipment with which we are endowed.
But Fodor's rejection of much past philosophical work should by no
means be interpreted as an uncritical return to the tenets of Cartesianism.
First of all, he stresses that mentalism is not equivalent to dualism. One
can believe in the existence of mental states and in their causal efficacy
without believing that there are two substances-mind and matter-which
must somehow interact with one another. In this case, he is putting forth
a materialistic variant of mind, which nonetheless allows the existence of
mental causes and the interaction of mental states with one another.
Indeed, Fodor accepts much of the functionalist perspective as first
introduced by his teacher, Hilary Putnam. On this view, the psychological
constitution of a system depends not on its hardware (or its physical
realization) but on its software: thus, Martians can have pains, and com-
puters can have beliefs. Fodor goes on to embrace the general information-
processing approach of the cognitive sciences: it is in the manipulation of
symbols, or mental representations, that cognitive activities are realized
and, in fact, constituted. Consistent with many contemporary philoso-
phers of mind, and in direct opposition to those earlier philosophers who
saw mind as a mirror of nature, Fodor rejects resemblance as a property
of mental representation: the symbols of the mind are abstract entities
which need bear no configurational relationship to the entities they denote.
Finally, Fodor's functionalism in no sense reflects a reductionist ap-
proach. Intelligence and mental states can be realized in many systems and
entities, and there is no priority given to explanation in terms of the
biochemical or the neurological. Indeed, Fodor voices his strong reserva-
tions that the "natural kinds" of the nervous system will map in any
interesting way onto the "natural kinds" of psychological or mentalistic
explanations. If anything, the tie between mind and computer is likely to
be more intimate than the tie between mind and brain. Throughout, there
is a strong "intentionalist flavor" to the discussion: Fodor injects an idiom
of beliefs, desires, goals, and the like entering into the very center of
discussions about various kinds of cognitive systems.
Until this point it may seem that, while Fodor features a different cast
of heroes and villains, his approach is consistent with those of other
cognitively oriented epistemologists, such as Putnam and Dennett. And,
indeed, viewed from afar, and compared with skeptics like Rorty or Ryle,
these philosophers can all be seen as enthusiasts of the computational
metaphor in the world of cognitive processes. But Fodor has gone well
beyond his contemporaries in his willingness to think about what mental
representation might be like. Here his Cartesian allegiance becomes most
patent.
82
Reason, Experience, and the Status of Philosophy
In a word, Fodor believes that there must be a language of thought.
If cognitive systems involve representations, if cognitive operations in-
volve the manipulation of symbol-like representations, then these repre-
sentations must exist somewhere and be manipulated in some way. Ac-
cordingly, Fodor feels that a commitment to attributing a representational
system to organisms must entail a characterization of this mentalistic sys-
tem. He asserts, "I am proposing ... to resurrect the traditional notion that
there is a 'language of thought' and that characterizing it is in good part
what a theory of the mind needs to do" (1975, p. 33). Fodor feels that
acknowledgment of some kind of medium, or language, within which
thinking takes place is an implicit part of nearly every contemporary
cognitive theory. It has become his burden to try to spell out what that
language might be.
And so to Fodor's long and highly stimulating essay The Language of
Thought (1975), seen by some commentators as a "great divide" in twen-
tieth-century philosophy (Piattelli-Palmarini 1983). In this work, Fodor
claims that the language of thought must be an extremely rich vehicle if
it is to carry out the many cognitive processes-perception, reasoning,
language-learning, and the like-of which human beings are capable. In-
deed, if mental processes are computational, there must be representations
upon which computations can be performed. Moreover, he puts forth his
radical view that this language of thought is innate: that people are born
with a full set of representations, onto which they can then map any new
forms of information that happen to emerge from their experiences in the
world. Furthermore, "the language of thought may be very [much] like a
natural language. It may be that the resources of the inner code are rather
directly represented in the resources of the codes we use for communica-
tion .... [This is] why natural languages are so easy to learn" (1975, p. 156).
Fodor's claim that individuals are born with knowledge of a language
-an innate language similar to a natural language-is astonishing, and it
is not certain just how seriously he means it to be taken. But his challenges
to competing accounts are deadly serious. Take, for example, his critique
of Jean Piaget's theory of concept acquisition, according to which the child
comes to possess new and more powerful concepts at each subsequent
stage of development. Fodor parades the difficulty he has in understanding
how one can ever learn a new concept unless one already has the ability
to hypothesize that concept-in which case, one already possesses it.
Shades of the Meno! Similarly, in his critique of the claim that complex
concepts are built out of simple ones, Fodor argues strenuously that all
attempts a!a Locke to identify the building blocks of larger concepts have
failed, and that this failure is putative evidence that this particular "em-
piricist" maneuver was ill conceived from the start. The fact that the
83
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
operations that individuals can carry out, even early in life, are highly
abstract, gives further weight to Fodor's (and Chomsky's) claims that the
initial intellectual apparatus with which individuals are equipped must be
sharply specified, constructed so as to expect particular experiences and
information. Even though the exact claims put forth by Fodor have failed
to persuade most of his colleagues, the issues he raises about the need for
some kind of "mentalese," and the kinds of constraints this mentalese may
have to exhibit, have proved difficult to undermine. Thus, his position
passes one of the most critical tests for any philosophical claim.
From Fodor's point of view, it is inadequate to conceive of the lan-
guage of thought as simply a formal medium of symbol manipulation (see
Stich 1983). There must be some way in which the contents of the world
are represented mentally, for we do not just think, we think about certain
things, and those certain things surely exist in the world. Yet at the same
time, Fodor expresses pessimism about the likelihood that we will ever be
able to understand how content is dealt with by our computational systems.
We may be restricted, as investigators, simply to describing the kinds of
"syntactic" operation that are carried out, while remaining ignorant of the
ways in which these operations refer to the external world so as to lead
to different mental concepts. As Fodor puts it, it may be impossible for the
cognitivist to do semantics, even though "to deny that mental operations
have access to semantic properties of mental representations is not to deny
that mental representations have semantic properties" (198lb, p. 244).
Fodor does fear that these properties may be inaccessible to scientific
investigation. The machine of our minds does not know what it is talking
about, and does not care about a semantic relation. While so-called natu-
ralistic psychology wants to be able to explain how we learn to know about
the particular things of the world and what they are like, computational
psychology-a psychology of formal mental operations-may be the only
kind of psychology we can ever hope to get. On such "content-blind"
sentiments the historically oriented student of philosophy can discern
links between Fodor's enterprise and that of Carnap and the logical empiri-
cists: indeed, Fodor even adopts Carnap's term methodological solipsism to
characterize a syntactic approach to cognition.
In sum, then, Fodor believes that any attempt to understand cognition
must involve a full-fledged embracing of a mentalistic point of view. He
believes that mental states really exist, that they can interact with one
another, and that it is possible to study them. Methods of study involve
the empirical methods of psychology, linguistics, and other cognitive
sciences; and chances of making advances on these issues are tied closely
to an informed collaboration among experts in these different areas.
Fodor's own guess is that those mental operations that occur in a relatively
84
Reason, Experience, and the Status of Philosophy
rapid and automatic fashion-like the syntactic parsing of a sentence or the
detection of forms in the visual world-are most likely to be elucidated by
current cognitive scientific methods. He is correspondingly pessimistic that
those capacities involving sustained judgment and reasoning-such as the
development of scientific theories or the making of decisions about ordi-
nary life-will lend themselves to the kind of syntactic (or formal) analysis
for which cognitive science has proved suited.
It might be thought that such a reversion to Cartesian notions, such
bold talk of a language of thought, would prove anathema to many of the
authorities I have discussed. In particular, it might be thought that Richard
Rorty would see Fodor's move as exactly the kind of tack a sophisticated
cognitive researcher ought to avoid. It is surprising, therefore, but also
illuminating, that while Fodor's work is reviewed critically in Rorty's
book, it is not by any means dismissed.
Rorty, it will be recalled, saves his harshest criticism for those who
believe in the Mirror of Nature-in a mental apparatus that in some way
can tell about what the world is like and help us judge it in the most
accurate possible way-to give us "right opinions," as he might put it.
Rorty notes that Fodor is interested in constructing a mechanistic model
of mental processes and is willing to posit all kinds of internal states for
doing so, but that his ultimate criteria call only for the kinds of description
that may yield a fruitful theory of cognitive processing-an avowedly
psychological non-epistemological program like Dodwell's. Rorty stresses
that the question, "How do we recognize bottles?," is entirely different
from the question, "What is indubitably given to the mind, such as to serve
as an infallible touchstone for inference?" (1979, p. 245). And so Rorty and
Fodor concur on the importance of separating out two questions: how the
organism interacts with the world-a legitimate psychological question;
and whether the organism's views about the world are in fact true-the
traditional (and, to Rorty's mind, the untenable) question. We now see,
Rorty points out, "that Fodor's picture of the mind as a system of inner
representations has nothing to do with the image of the Mirror of Nature
I have been criticizing. The crucial point is that there is no way to raise the
skeptical question 'How well do the subject's internal representations rep-
resent reality?' about Fodor's 'language of thought'" (p. 246).
Whatever Fodor's and Rorty's particular points of agreement may be,
Fodor, Chomsky, and others in the cognitive camp clearly wear the mantle
of Rorty's systematizers: they believe that the traditional questions are
important, and they want to try to solve them as best they can, though
they may retain rather more skepticism about what is possible than did the
great systematizers of a few centuries past. Rorty, on the other hand, as
a self-styled edifier believes that philosophy has, at most, puzzles, and that
85
I I I THE CoGNITIVE SciENCEs : HISTORICAL PERSPECTIVE
86
Reason, Experience, and the Status of Philosophy
don these issues-leaving them to empirical scientists to work out if they
so chopse-and instead become informed commentators about culture-
which in the case of our culture would include commenting on the great
philosophers of the past. A different argument is put forth by researchers
in the mainstream of cognitive science, for example, in artificial intelli-
gence. From their perspective, once one has provided computational ac-
counts of knowledge, understanding, representation, and the like, the need
for philosophical analyses will evaporate. After all, philosophy had once
helped set the agenda for physics; but now that physics has made such
tremendous strides, few physicists any longer care about the musings of
philosophers (Holton 1984).
I am not convinced by either of these critiques. On my analysis,
philosophy has been crucial from the first in formulating a set of questions
worthy of study and in monitoring the course of these questions over the
millennia. There is progress in philosophy-not perhaps as linear in the
physical sciences-but clear progress nonetheless. This progress comes
about because of the debates that take place among philosophers-for
example, when Locke criticized certain concepts of Descartes, arid Berkeley
and Hume in tum criticized Locke; and it occurs with equal impact through
interaction between philosophers and empirical scientists, such as those
discussions in which physicists and logical empiricists were involved in the
early part of this century.
I suggest, then, that philosophy participates in the disciplinary matrix
by virtue of its dialectical role: a dialectic within the discipline and a
dialectic between the analysis put forth by philosophers, on the one hand,
and the empirical findings and theories put forth by scientists, on the other.
This has happened dramatically in recent years. Just at the time when
philosophy seemed at low ebb, when the program of logical empiricism has
been thoroughly discredited, the invention of the computer and the begin-
ning of cognitive science suddenly underscored the need for sophisticated
analysis. It was thinkers acquainted with the long-standing philosophical
tradition-with Kant's notions about representations, Descartes's claims
about the mind-body problem, Locke's skepticism about innate ideas-
who could bring to bear the appropriate conceptual framework and then
revise it in the light of new scientific discoveries.
All this is not to claim that the traditional questions of philosophy
have inherent superiority. While Rorty's critique of traditional epistemol-
ogy seems like "overkill" to me, the kinds of questions pondered by
philosophers have changed over the millennia: some issues, such as the
nature of visual perception, which exercised the Greeks, become the exclu-
sive province of empirical science; some issues, such as the nature of "raw"
sensory experience, come to be seen as ill conceived; and some issues, like
87
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
the nature of intention and purpose, gain new urgency because of inven-
tions like the computer.
I see the invention of cognitive science as a wonderful stimulus for
philosophy, on the one hand, and philosophy as an indispensable hand-
maiden for the empirical scientists, on the other. Philosophy enables us to
define fundamental cognitive scientific questions in a coherent way, and
assures the proper integration of work in disparate fields. But, by the same
token, philosophy must attend assiduously to empirical findings in order
to avoid becoming a barren discipline or one irrelevant to scientific work.
It is thus fitting that the field of philosophy, whose initial agenda helped
to stimulate the rise of cognitive science, has been fueled by that new
discipline, even as philosophy can, in tum, help to inform and interpret
work spawned by its recent intellectual offspring.
Hilary Putnam, a veteran of many of these discussions, has reflected
on the role of philosophy in the contemporary scientific scene. His prudent
comments take seriously the various critiques of "grand" philosophy,
while recognizing the important role philosophy should continue to play
in discussions of new scientific endeavors:
I have not attempted ... to put forward any grand view of the nature of
philosophy; nor do I have any such grand view to put forth if I would. It will be
obvious that I do not agree with those who see philosophy as the history of
"howlers" and progress in philosophy as the debunking of howlers. It will also be
obvious that I do not agree with those who see philosophy as the enterprise of
putting forward R priori truths about the real world .... I see philosophy as a field
which has certain central questions, for example, the relation between thought and
reality .... It seems obvious that in dealing with these questions philosophers have
formulated rival research programs, that they have put forward general hypothe-
ses, and that philosophers within each major research program have modified their
hypotheses by trial and error, even if they sometimes refuse to admit that that is
what they are doing. To that extent philosophy is a "science." To argue about
whether philosophy is a science in any more serious sense seems to me to be hardly
a useful occupation .... It does not seem to me important to decide whether science
is philosophy or philosophy is science as long as one has a conception of both that
makes both essential to a responsible view of the real world and of man's place in
it. (Putnam 1975R, p. xvii)
88
5
Psychology:
The Wedding of
Methods to Substance
89
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTivE
90
Psychology: The Wedding of Methods to Substance
ist approach-had eventually foundered. In recent years, the most exciting
new work in the human sciences had come from two connected areas:
information theory, which posited principles of transmission applicable to
any kind of channel; and computer science, which now featured machines
engaged in symbol manipulation. Miller was holding out hope of marriage
between the quantities of data collected by psychologists over the years
and the rigorous new approaches of the engineering-oriented scientists.
The result might be a genuine science of psychology with its own set of
immutable laws. No one thought to question whether all contents, or bits,
can in fact be treated (and then counted) as equivalent.
91
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
intimately linked to George Miller's studies of the magic number 7.) There
was an important added twist: rather than simply speaking of structural
limits in a static way, the British researchers sought to determine precisely
what happens to this information from the moment one first apprehends it.
Given this "engineering" approach, it became a natural step to draw a flow
chart of what happens, as the perceptual system operates upon new infor-
mation. In fact, according to a recent textbook, Broadbent was the first
psychologist in modern times to describe cognitive functioning with a flow
chart (Lachman, Lachman, and Butterfield 1979, p. 188).
Store of Conditional
Probabilities of
Past Events
What was this early flow chart like? It featured information coming
in through the senses, being placed in a short-term store, and then being
selectively filtered before entering into a limited-capacity perceptual sys-
tem. While a sense organ can take in a lot of information in parallel and
retain it momentarily, the job of the selective filter is to block unwanted
messages, thereby letting in only those that merit additional analysis. A
further property of the selective filter is that it can be so tuned that, at any
one time, it allows in only those messages that fulfill certain requirements.
The buffer can hold unanalyzed information briefly and thus allows one
to report the contents of the second ear, after having spewed out the three
digits apprehended by the first ear. According to Broadbent's early model,
only information that becomes conscious-that passes through the limit-
92
Psychology: The Wedding of Methods to Substance
ed-capacity channel-can enter into long-term memory and thus become
part of active knowledge. Information present on an unattended channel,
or in an unattended signal, is assumed to decay in a few seconds and to
receive no processing beyond the initial "pre-attentive" analysis.
While few, if any, investigators believed that perception or thinking
takes place simultaneously or without a series of steps, the option of
tracing the stages of information processing had rarely been followed
before Broadbent's time. Again, this option became a probability when
communication engineering began to impinge on the perceptual and atten-
tional issues that had long interested psychologists. But the model of "flow
charting" put forth by Broadbent and his associates and their evidence
relevant to specific stages of information processing opened up many pro-
ductive possibilities. One could now examine the temporal dimensions of
diverse psychological processes, and avid experimenters lost little time in
pursuing just that course. The lack of attention to the particular content
being processed, or to the lcinds of transformation imposed, did not trouble
those excited by the Broadbent-Cherry demonstrations.
93
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
the past few years have witnessed a notable increase in interest in and investi-
gation of the cognitive processes.... It has resulted from a recognition of the
complex processes that mediate between the classical "stimuli" and "responses"
out of which stimulus-response learning theories hoped to fashion a psychology
94
Psychology: The Wedding of Methods to Substance
that would by-pass anything smacking of the "mental." The impeccable peripher-
alism of such theories could not last.... One might do well to have a closer look
at these intervening "cognitive maps." (P. vii)
The importance of Bruner's book was signaled by praise from the distin-
guished physicist J. Robert Oppenheimer: 'A Study of Thinking has in many
ways the .flavor of the opening of a new science.... The book has a unity
of view and a fervor of conviction which makes it point to the future"
(quoted in Bruner 1983, p. 121). The possibility that the use of such
artificial concepts might invalidate the findings was far from anyone's
mind at that time.
95
II I THE CoGNITivE SciENCES : HISTORICAL PERSPECTivE
tation, and of allowing subjects to use their full reflective powers was
bracing and freeing.
Psychology is a discipline central to any study of cognition. Yet it is
also a difficult discipline to pursue and one where genuine progress has
been hard to achieve. Nearly every conceivable element is relevant to a
subject's performance, and few issues having to do with human nature and
behavior can be excluded from the laboratory a priori. Thus, choosing a
problem, and screening out all competing ones, becomes an especially
vexing task.
Psychology also poses special problems for the historian of cognitive
science-a problem in no way minimized when the historian is also a
psychologist. It is an enormous field-there are many more psychologists
than there are representatives of the other fields-and there are conse-
quently more programs of research to survey. While it is oversimplification
to organize any field around one or two themes, it is especially difficult to
select key issues in psychology. Should one, thus, pay attention to the
particular content of information (auditory or visual, musical or linguistic)
or instead treat all contents as if they were interchangeable? Does one
approach research in order to illuminate those processes that are true of all
individuals, or look instead at pertinent individual differences-child ver-
sus adult, male versus female, naive versus trained in experimental tasks?
Does one examine behavior in its natural context or try to strip away all
everyday accouterments and resort to artificial laboratory conditions?
Does one assume that the individual approaches tasks by building up
larger elements of meaning from small, isolated units? Or does it make
more sense to assume that one comes to tasks with general strategies or
scripts, which one simply imposes upon a task, irrespective of its particular
dimensions, details, and demands?
I have elected to organize this chapter in terms of a distinction that
touches upon some of the aforementioned ones but is perhaps better
phrased in somewhat different terms: that is, the distinction between
molecular or small-scale units of analysis and molar or large-scale units of
analysis. For reasons of scientific strategy or simple personal preference, it
seems possible to classify most psychological research programs along this
dimension. Some programs, such as those of traditional psychophysics and
contemporary information processing, show a penchant for small-scale
units (bits, individual percepts, single associations examined in brief peri-
ods of time) on the assumption that a thorough understanding of these
elementary units and processes is the surest path toward the ultimate
explanation of complex units and entities. A contrasting faith is found
among proponents of the molar approach-those who look at large-scale
problems tackled over a long period of time and invoke analytic concepts
96
Psychology: The Wedding of Methods to Substance
like schemas, frames, or strategies. According to these researchers, these
large-scale properties are most salient in human cognition and thus serve
as a logical point of departure. Why gamble that an elementaristic ap-
proach will eventually yield larger units, when one has the option of
beginning instead with these larger units, which seem closer to the data and
the experiences of everyday life?
The contrast between molecular and molar approaches resembles, but
is by no means identical to, the distinction between the top-down and the
bottom-up approaches. In a top-down approach, which has rationalist over-
tones, the subject is assumed to bring to the task his own schemes, strate-
gies, or frames which strongly color his performance. In a bottom-up
approach, more allied to the empiricist camp, the actual details of a focal
task or situation are assumed to exert primary influence on a subject's
performance. In what follows, I shall often identify molar with "top-down"
and molecular with "bottom-up"-not because each is logically bound to
the other but because they often, and perhaps typically, occur together.
Like all dichotomies, this one is easily exaggerated, with subsequent
distortion of the field. Nearly all psychologists have some sympathy for
each tack, and many move from a molecular to a molar approach (and back
again). For example, George Miller favored a molecular hat when ponder-
ing the number 7 but readily shifted to a molar one when discussing plans
and goals in the 1960 volume. Indeed, when the computer is used as a
model, it is equally justifiable to focus on the most molecular level (indi-
vidual bits, symbols, "on-off" circuits) or the most high-level program-
ming concepts (goals, means, and routines). Also, one can embrace a mo-
lecular (or a molar) approach for different reasons: some psychologists
begin with a molecular approach in the hope of being able to adapt their
methods to molar entities; while others believe that ultimately all behavior
can be reduced to, and explained by, molecular entities. Thus, in embracing
this dichotomy, I seek to convey an ongoing tension or struggle for the soul
of psychology-not to label two bins into which one can readily and
reliably sort all experiments, concepts, and psychologists.
Two other trends must be mentioned as well in any thumbnail
sketch of psychology's first one hundred years. The first trend is the
increasing splintering of the field. The American Psychological Associa-
tion alone has over fifty thousand members (including several thousand
active researchers) who spread themselves over forty divisions and sev-
eral hundred special interest groups, many of which are completely igno-
rant of what is going on elsewhere in their association and their disci-
pline. In this climate, efforts to find unifying concepts are vital but by no
means easy to sustain.
The second is the trend toward methodological perfection. With the
97
I I I THE CoGNITIVE SciENCEs : HISTORICAL PERSPECTIVE
98
Psychology: The Wedding of Methods to Substance
subject matter difficult, if not impossible, to bring to bear on complex and
all-pervasive human interaction.
Kant's most severe objections derived from his overall conception of
the domain of knowledge. In his "foundational" view, it was the province
of philosophy to lay out the nature of thinking-to chart the relations
among the various sciences and to designate their foundations and limita-
tions; psychology was seen as a second-rate poacher upon such a program.
Psychology should be content to look at the social and historical contexts
in which thinking occurs but should not attempt to crack the nature of
thought itself.
Such was the authority of Kant-and the surface persuasiveness of his
arguments-that many scholars of his time shied away from the empirical
investigation of psychological issues. Fortunately, however, at least some
individuals were spurred by Kant's skepticism to seek a more positive role
for psychology. According to the historian of psychology, David Leary
(1978), a trio of German thinkers-Jakob Friedrich Fries, Johann Friedrich
Herbart, and Friedrich Eduard Beneke-directed their energies, in the early
part of the nineteenth century, to the conceptualization of a scientific
psychology. Each of these scholars believed that mental processes could be
measured by experiment and that studies could be carried out that actually
provided information on the operations of the mind. In particular, Herbart
claimed that ideas exhibit the variables of time, intensity, and quality, and
that one should be able to measure quantitatively each of these aspects of
ideation and even to write equations mapping their relations one onto
another. While the work of Her bart and his colleagues was basically re-
stricted to armchair speculation, these scholars kept alive the possibility of
a scientific psychology during an era where Kant's strictures remained a
formidable obstacle.
By the middle of the nineteenth century, scientists came to have fewer
reservations about empirical investigations bearing directly on psychologi-
cal issues. Less under the shadow of Kant and other philosophers, more
willing to carry out experiments and simply see what would turn up, these
scholars directly anticipated the founding of scientific psychology toward
the end of the century and set up ripples that can still be detected in the
laboratories of today. Moreover, in each case, they made specific contribu-
tions that remain pertinent to psychological discussions.
99
I I I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
show that much of what Kant had speculated about could be subjected to
empirical study. Skeptical of various competing claims about the astound-
ing speed of thought, Helmholtz actually undertook to measure the length
of time it takes to transmit impulses along a nerve. Using an ingenious
instrument, adapted from his laboratory galvanometer, Helmholtz was
able to measure the time that it took for a nerve impulse to pass through
the severed leg of a frog. He then succeeded in adapting the method for
human subjects, who were asked to respond by pushing a button when-
ever a stimulus was applied to their legs. This allowed him to ascertain the
speed of the impulse in human sensory nerves: it turned out to be between
165 and 330 feet per second (Fancher 1979, pp. 100-101). Human behav-
ioral reactions could be measured, after all.
Having successfully challenged the belief that thought is essentially
instantaneous or inherently unmeasurable, Helmholtz went on to call into
question Kant's beliefs in innate ideas of space. According to Helmholtz's
rival position, individuals build up knowledge of space, just as they con-
struct the facts of their physical world. By having subjects wear prisms
which distorted their world, Helmholtz showed that individuals could
readily adjust or adapt to these distortions and were soon seeing the world
again in essentially undistorted fashion. In a complementary investigation,
he studied individuals who had once been blind but could now see, and
documented the considerable time that elapsed before they learned to
perceive the world of objects in the manner of the sighted.
From such demonstrations, Helmholtz developed the still-influential
idea of unconscious inference: rather than simply reading off percepts from the
world of external stimulation, we unconsciously draw on our past knowl-
edge in order to effect accurate interpretations of what we perceive. The
experience of one's past perception is unconsciously added to one's present
reaction to a stimulus-as happens when, defying the evidence of the
senses, one succumbs to an optical illusion. Helmholtz used the word
inference deliberately. He believed that the visual system was implicitly
reasoning about its experiences: for example, in order to figure out the
actual sizes of objects, the visual system had to make inferences based on
the images formed on an individual's retina. Unlike syllogistic inferences,
of course, these processes occurred without conscious awareness.
In forging ahead with his research, and in introducing certain concep-
tual distinctions, Helmholtz made three major contributions. First of all,
he indicated that Kant's philosophical dicta did not have absolute validity:
it was indeed possible to illuminate aspects of human mental functioning
in an empirical fashion. Second, Helmholtz cleared places for molecular
forms of analysis (the speed of an impulse traveling along a nerve fiber)
as well as molar investigations (the ways in which complex spatial arrays
100
Psychology: The Wedding of Methods to Substance
are seen under both normal and distorted conditions). Finally, by stressing
the perceiving subject's contribution to perception, Helmholtz became an
early contributor to the ideology of cognitive science.
It now became possible for scientists of a less overpowering status to
contribute to the incipient science of psychology. The pioneering psycho-
physicist Gustav Fechner (1912) was able to show that, within limits, the
intensity of a perceived sensation varies as a logarithmic function of objec-
tive features of the stimulus. Thus it seemed that the most personal aspects
of a psychological experience-how loud a sound seems, how bright a light
appears, how sweet an object tastes-could bear a quantitative relation to
a measurable characteristic of an object in the world.
Building on Helmholtz's documentation of the speed of neural events,
F. C. Donders proposed in 1868 that one could also measure the time that
it takes for higher mental operations to be carried out. For instance, in order
to measure how long it takes to effect discrimination between two stimuli,
one would subtract the time that it takes to detect a single event in isola-
tion from the time it takes to respond to only a single event when two
events are presented. By subtracting the time needed to detect a single
event in isolation from the time needed to respond when given two stimuli,
one could infer the exact time required for the operaHon of discrimination. Like
Fechner, Donders worked with relatively molecular materials: but, in prin-
ciple, their methods could be applied to molar tasks.
While Helmholtz, Fechner, and Donders were all concerned to dem-
onstrate that psychological issues could yield quantitative results in the
experimental laboratory, a priest-philosopher named Franz Brentano was
approaching questions of psychology from a different standpoint. Bren-
tano's writing was directed against the molecular notion that one can break
psychology down into elements and examine the elements of experiences
or consciousness in isolation, as well as the notion that one can conceive
of thought processes or consciousness in a purely mechanistic way. For
Brentano, psychology starts with the mind-an active, creative entity
which has intentions, for it implies and demands an object. The true
subject matter of psychology is the mental act-such as judging, sensing,
imagining, or hearing, each of which reflects a sense of direction and
purpose. One cannot simply see; one must see something; and the act of
seeing something is psychological or mental.
Given this viewpoint, the task of empirical psychology is to study the
mind of the agent at work, dealing with objects, purposes, and goals.
Brentano stressed the phenomenological aspects of the psychological en-
deavor: one cannot conceive of thoughts and judgment, let alone study
them, except by taking into account one's inner phenomenal experience.
And this can be accessed not by prompted introspection-for one cannot
101
II I THE COGNITIVE SCIENCES : HISTORICAL PERSPECTIVE
observe at the same time that one experiences-but rather by simple phe-
nomenal experience of one's inner mental life (Brentano 1874).
The words of Brentano would be stilled temporarily during the first
years of the discipline of psychology, when an obsession with studying the
elements of sensations, and the basic constituent mental states, carne to the
fore. The laboratories of Leipzig had little place for such high-flown senti-
ments as act, purpose, intention, and phenomenal experience. Yet the
kinds of top-down or molar concern that Brentano had raised could not
forever be ignored. They were to re-emerge, in different forms, in the
rebellion in Wiirzburg, in the platform of the Gestalt psychologist, and,
once again, in the view of the computer as an agent with plans, intentions,
and goals. In not a few ways, Brentano has had profound influence on the
course of psychology and has affected many experimenters who have
never heard his name.
Wundt s Program
In a sense, psychology was well under way by the latter part of the
nineteenth century. With the pioneering work of Helmholtz, and the con-
siderable efforts by Donders, Fechner, and Brentano, who devoted them-
selves chiefly to psychological matters, there was already a minor industry
of psychological thinking and research. And yet, as I shall note repeatedly
in this history, the flowering of a discipline depends heavily upon the
founding of institutions and organizations. The efforts of the earlier nine-
teenth-century pioneers might never have coalesced and become cumula-
tive had it not been for Wilhelm Wundt, who took it upon himself in the
latter part of the century to establish psychology as a separate experimen-
tal discipline. More than any other individual-perhaps than any collec-
tion of individuals-Wundt is responsible for the emergence of psychol-
ogy as a separate scientific discipline with its own methods, programs, and
institutions. Psychology has grown from its original establishment in a
single university-Leipzig in 1879-to a field with representatives in virtu-
ally every institution of higher learning. Its spectacular success can be
attributed in no small measure to the labors-if not always to the ideas-
of Wundt (Boring 1950; Fancher 1979; Watson 1979).
According to Wundt, physics studies the objects of the external world:
while this investigation is necessarily mediated by experience, physics is
still not the study of experience itself. Psychology, in contrast, is the study
of conscious experience as experience. It must be approached through
internal observation, through introspection. While all individuals have
such experiences, not all are necessarily qualified as expert witnesses on the
nature of their experience. Thus Wundt embraced the method of intro-
spection-a method whereby one attends carefully to one's own sensations
102
Psychology: The Wedding of Methods to Substance
and reports them as objectively as possible. Such objectivity here means
that one describes the sensations felt, rather than the stimulus giving rise
to them; and that one reports thoughts (or images) without reference to
their meaning or context of presentation. Wundt's program hinged on the
possibility of introspecting in this fashion.
Just as it was important to separate psychology from physics, it was
also important to effect its divorce from physiology. Physiology has as its
special province the study of mental processes that are not conscious-not
susceptible to introspective examination-while psychology appropriates
those higher mental processes that could be subjected to personal examina-
tion. Physiology explains how we have sensations, but psychology covers
the description and analysis of sensory experience. Recognizing that some
aspects of human experience prove less hospitable to examination in terms
of introspection, Wundt made a further discrimination within psychology
between those individual experiences that are susceptible to introspection
and those aspects of human experience that are by their nature social or
communal. In Wundt's division of labor, it fell to ethnic or folk psycholo-
gists to study these complex and large-scale human activities, such as
customs, rituals, and, perhaps, certain features of language and thought.
Ultimately, he contributed ten volumes to this effort at constructing folk
psychology.
As George Miller has put it, Wundt looked at the agenda of British
empirical psychology with an eye of a man trained in the traditions of
German physiology (Johnson-Laird and Wason 1977, p. 2). Like Helmholtz
and Donders, Wundt devised simple tasks-but asked subjects to intro-
spect about them during the course of their participation. In one sample
task, subjects were asked to press a button whenever they heard a tone.
On some trials, the subjects' attention was directed to the tone; while on
others, their attention was directed to the movement of their fingers on the
button. It turned out that reaction times were longer when a subject had to
attend to the sound, shorter when one attended to the movement. Wundt
inferred that this greater time was due to apperception-the process of mak-
ing one's experience clear in one's consciousness. Apparently, when one
is concentrating on a stimulus, the stimulus first has to be perceived and
then apperceived, or consciously interpreted, in the light of the response
associated with it. Even further complications resulted when several differ-
ent stimuli were presented, only one of which was targeted for response.
This process took even longer, because, according to Wundt, "cognition"
has to occur: the stimulus has not only to be perceived and apperceived
but also differentiated from other stimuli that are not to elicit responses
(Fancher 1979, p. 139).
Wundt revealed his ties to the British philosophical tradition in the
way in which he made sense of such findings. He came to think of experi-
103
II I THE CoGNITIVE SciENcEs : HISTORICAL PERSPECTIVE
104
Psychology: The Wedding of Methods to Substance
Hermann Ebbinghaus (1913)-simply went their own way. Inspired with
the fertile idea of ferreting out principles of memory through the use of
materials that could not be contaminated by earlier experiences and as-
sociations, Ebbinghaus generated over two thousand nonsense syllables
and measured his own skill at learning sets of them. Practicing them either
at one sitting or over a series of sessions, Ebbinghaus probed the effects
of drill time between presentations, backward and forward interference,
and other "independent variables." His stimuli were no more meaningful
than Wundt's; but, by breaking away from introspection as the sole source
of information, and by instructing novel methods of statistical analysis,
Ebbinghaus succeeded in reorienting psychology in a very productive
manner. It was one's actual skill at a task-not one's introspections about
what one senses or feels-that became the object of study and measure.
Ebbinghaus's methods have exerted far more of an impact on experimental
psychology than has Wundt's introspective tack. Yet, influential as this
work has been, the question must be raised whether the immaculate,
content- and context-free approach of Ebbinghaus ultimately accom-
plished more harm than good; for we now know that methods used to
remember ordinary events differ in important particulars from meth-
ods used to remember meaningless arrangements of letters or organized
material.
105
II I THE CoGNITIVE SciENCES : HisTORICAL PERSPECTIVE
But in its more general form, Wiirzburg had launched a severe critique of
the entire manner in which Leipzig went about its investigations and drew
its conclusions.
The story begins innocently enough in 1901, when the young psy-
chologist Karl Marbe asked his subjects to compare weights with one
another-a familiar procedure, dating back to Fechner's first reveries.
Marbe found that, in judging weights, subjects reported no imaginal con-
cepts as the basis of judgment: in other words, no images flitted through
their introspecting minds. Instead, counter to the Wundtian expectation,
subjects reported that various vague attitudes passed through conscious-
ness-attitudes like hesitation, doubt, waiting for an answer, feeling that
the answer had arrived. Marbe was forced to conclude, "The present data
are quite sufficient to draw the conclusion that no psychological conditions
of judgments exist .... Even ... the observers concerned ... were extremely
surprised to note the paucity of experiences that were connected with the
judgmental process" (Mandler and Mandler 1964, p. 143).
That some subjects failed to report imagery while rendering a judg-
ment may hardly seem like headline news in the scientific community, but
it ran against the widespread dictum that all thought features imagery
accessible to consciousness. Then, in the following years, a whole raft of
studies emanating from Wiirzburg repeated and elaborated upon this ini-
tial negative report (see Boring 1950; Humphrey 1951; Mandler and Man-
dler 1964). There was Henry Watt who reported that the conscious task
(or Au/gabe) that a subject was posed had an important effect on the kinds
of associations one made. There was Narziss Ach who reported the opera-
tion of a determining tendency: the task had the effect of orchestrating various
associations and skills into a purposeful orderly sequence-a kind of "di-
recting" will-which led smoothly to its final execution. There was August
Messer who saw consciousness as the visible portion of an iceberg, with
most thought processes occurring beneath the surface, and with conscious
processes themselves exhibiting varying degrees of clarity. And there was
Karl Buhler, one of the seminal figures of psychology, who dismissed the
simple problems posed by the Leipzigers and posed truly complex tasks to
his subjects-for example, the discussion of philosophical problems. Like
his Wiirzburg colleagues, Buhler uncovered peculiar varieties of conscious-
ness like doubt, astonishment, and even consciousness of consciousness.
Oswald Kiilpe sought to put these different demonstrations together.
In a critique of the Wundt-Leipzig program, he pointed out that it was not
enough merely to pose problems to individuals and let them introspect as
they wished. It was necessary to pose problems making different kinds of
demand and to monitor their varying effects. One could not simply assume
that all important aspects of mental processes are conscious, that images
106
Psychology: The Wedding of Methods to Substance
always guide thinking, and that mental contents (like sensations) are nec-
essarily the constituent elements of thought. There was a more positive
aspect to the critique as well. Thanks to the work of Kiilpe and his associ-
ates, mental acts like attending, recognizing, willing, comparing, and differ-
entiating came to constitute a proper sphere of psychology. While, to be
sure, these acts lack vivid perceptual characteristics of sensations, images,
and feelings, they are no less important on that account. Most generally,
there was an increased willingness to recognize top-down "structuring
tendencies" in the solution of these problems and to use problems that
truly engaged a subject's ratiocinative capacities. Here is a harbinger of the
molar concerns of many contemporary cognitive scientists.
The Wundtians did not accept such criticism without a rejoinder.
Points were scored on both sides-Wundt's more methodological, Wiirz-
burg's (to my mind) more substantive; but in the end, this disagreement
had a radical consequence. It called into question the merits of any psychol-
ogy that relies heavily on introspection, and particularly on the introspec-
tions of trained subjects, an uncomfortably high proportion of whom are
drawn from the ranks of the experimenters themselves. As vast an enter-
prise as a new science could not countenance so vague and subjective a set
of basic procedures.
107
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
108
Psychology: The Wedding of Methods to Substance
William James (and also by James Angell and John Dewey), a far more
radical shift was to occur in American psychology. This shift to behavio-
rism may initially have been necessary in order to put firmly to rest the
excesses of structuralism and introspection, whether in their Leipzig or
their Wiirzburg guise. Yet, from the point of view of a history of cognitive
science, it is difficult to think of this phase as other than primarily negative
and regressive.
109
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
110
Psychology: The Wedding of Methods to Substance
nitivism has prevailed for some of the reasons behind behaviorism's origi-
nal success: an exciting and refreshing new approach was seen as supplant-
ing a time-worn tack mired in questions of little interest to anyone outside
the field. But despite behaviorism's relatively rapid decline, only a few
lines of work in psychology fall largely outside the behaviorist camp.
Among these, I shall focus on the most direct link between the cognitively
oriented psychology of 1900 and that of today-the school of Gestalt
psychology.
Origins
Gestalt psychology can be introduced as a school rooted in several
dramatic demonstrations, which make a vivid point and thereby inspire a
certain way of thinking about mental phenomena. The initial demonstra-
tion of Gestalt phenomena was made in 1890 by Christoph von Ehrenfels,
an Austrian student of Brentano's. Ehrenfels's particular interest was the
perception of melody. He argued that the perceptual"form quality" in-
volved in a melody cannot be properly viewed as simply the sum of its
several tonal elements: indeed, it is an overall quality, a Gestalt, that
transcends its particular elements. As he pointed out, one could take the
same set of elements, or tones, and produce an entirely different melody.
Conversely, one could select an entirely different set of tones-for in-
stance, those in another key-and produce a figure that would be ap-
prehended as the "same melody" as the original one.
Ehrenfels is generally regarded as a predecessor, rather than the
actual founder, of the Gestalt movement. That honor belongs to Max
Wertheimer, who in 1912 published a paper on the visual perception of
movement. Working with two young assistants, Wolfgang Kohler and
Kurt Koffka, Wertheimer carried out a set of studies on apparent motion,
or the "Phi" phenomenon-roughly, the perceptual experience of move-
ment which arises when a set of lights or forms appears one right after
another (as in neon billboards or, for that matter, motion pictures). Wer-
theimer did not discover apparent motion-the phenomenon had been
recognized for some time; but he showed that the standard psychological
account of this phenomenon was untenable.
In particular, Wertheimer ruled out the common hypothesis of the
perception of movement being due to eye movements: he showed that
111
I I I THE COGNITIVE SCIENCES : HISTORICAL PERSPECTIVE
112
Psychology: The Wedding of Methods to Substance
"restructuring" of the elements in a situation. Thus, in order to fetch a
banana that was out of reach, an ape would have to move a chair from one
site to another, join two sticks together, or engage in some other reorgani-
zation of the situation that it beheld. Kohler (1925) found standard trial-
and-error accounts inadequate to account for the apes' behavior. Instead,
what seemed to happen was far more of a humanlike thought process,
where the ape would stop, reflect, and then, as if struck with a sudden flash
of insight, reach for the chair or the rope that provided the solution.
According to Kohler's analysis, the chimpanzee "restructured" the field
that had been presented to it; and did so courtesy of a moment of insight
-what the Wiirzburg school had termed an "A-ha experience."
Kohler also distinguished between bright and stupid apes. While a
bright ape would regularly exhibit these moments of insight, the dull ape
worked in quite a different fashion. Even when it saw the correct behavior
modeled, the ape would not be able to gain the treasure. Instead, in piece-
meal fashion, the ape would imitate the component actions without ever
apparently appreciating how to link them in order to secure a reward.
According to Gestalt psychology, the most primitive forms of learning
can be explained in terms of mere repetition or piecemeal associations. In
contrast, what characterizes higher learning or "intelligent" processes,
wherever found, is the capacity to grasp the basic fundamental relations
in a situation. These are the criteria of insight: "the appearance of a com-
plete solution with reference to the whole layout of the field" (Mandler
and Mandler 1964, p. 248).
Other scholars in the Gestalt tradition extended this line of study to
problem solving in human beings. Wertheimer (1945) himself examined
the solution of geometry problems, arithmetic puzzles, and even the steps
through which Einstein allegedly passed in arriving at his theory of relativ-
ity. Karl Duncker (1945) focused on the solution of engineering problems
-for example, the way in which one can direct X rays to kill a tumor
without destroying an unnecessary portion of intervening tissues.
Abraham Luchins (1942) examined the phenomenon of functional fixed-
ness: the way that the customary uses to which a material are put can
inhibit an individual (ape or professor) from perceiving how to use that
same implement in a novel way in order to solve a problem.
Note that the kinds of molar problem posed stood in sharp contrast
to those molecular tasks typically posed by the early German structuralists
and the American behaviorists. These molar problems are somewhat closer
to those posed by the Wiirzburgers but unencumbered by the weight of
introspections and elaborate interpretive terminology, and are forerunners
of problems favored today by many researchers in artificial intelligence.
The modes of explanation are also of a different order. Wertheimer and the
113
I I I THE CoGNITIVE SciENcEs : HISTORICAL PERSPECTIVE
114
Psychology: The Wedding of Methods to Substance
ing, a solitary psychologist working in Britain was also keeping the cogni-
tive faith. In his own explorations of memory, Frederic Bartlett had sought
to use, but had found inadequate, the strict experimental methods pio-
neered by Ebbinghaus. They seemed to miss much of what was central in
remembering meaningful content, and Bartlett came to the conclusion that
it is not possible to employ totally arbitrary materials and still capture the
salient features of memory on the wing. A fundamentally different ap-
proach was wanted.
Bartlett had vague intimations that memory was more of a social or
cultural phenomenon, and that the "set" accompanying a stimulating ex-
perience has a crucial effect on what one remembers and how well one
remembers it. Yet, according to his own recollections, Bartlett was stymied
for an experimental approach until he had a conversation with his friend
Norbert Wiener. In an important moment in the intellectual history of
cognitive psychology, Wiener gave Bartlett a crucial idea, as the latter
recalls:
115
I I I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
sense) by Kant and introduced into the psychology of the period by the
British neurologist Henry Head, Bartlett tried to make sense of his
findings in terms of the notion of schemas (sometimes called "schemata").
Bartlett claimed that the typical memory system used by humans in-
volves the formation of abstract cognitive structures, or schemas. These
schemas arise from prior encounters with the environment, as a result of
which certain kinds of information have come to be organized in spe-
cified ways. Thus, in listening to "The War of the Ghosts," subjects
would draw on their schemas for dealing with daily experience, in gen-
eral, and for dealing with adventure or ghost stories, in particular. To
make sense of this story of a war party proceeding on a canoe in order to
kill people, the listeners would use experiences of their own-say, a
canoe trip at night on a river-as well as structured information from
earlier stories-a typical adventure story involving primitive folks and
ghosts. To the extent that the information in Bartlett's story was consist-
ent with these previously constructed schemas, recall would be aided and
might tum out quite accurate. On the other hand, divergences between
the prior schemas and the details of the present story would cause sys-
tematic distortions in the initial recollection of the story and probably
introduce even further deviations in subsequent retellings.
Speaking more generally (and no doubt with the Ebbinghaus non-
sense-syllable tradition in mind), Bartlett declared:
Bartlett did more than keep alive a model for molar studies; he directly
anticipated the self-reflective system that cognitive scientists like George
Miller view as central to human cognition.
116
Psychology: The Wedding of Methods to Substance
biological principles, he decided to take a brief detour to study the devel-
opment of thought in children. As he liked to say, this detour lasted a
lifetime.
In the course of a sixty-year career, rivaling Wundt's in productivity
and exceeding Wundt's in its influence upon subsequent research, Piaget
provided portraits of the developmental course of children's thought, in a
variety of domains (see Piaget 1970 for an overview). Nearly always, his
work exhibited two hallmarks. On the one hand, conforming closely to my
definition of a cognitive scientist, Piaget took as his research agenda the
great issues of Western epistemology: the nature of time, space, causality,
number, morality, and other Kantian categories. He regarded these catego-
ries not as givens in the mind but rather as categories to be constructed,
Helmholtz style, over the course of a child's development. On the other
hand, Piaget insisted on painstakingly careful observations of children:
sometimes as they were engaged in free exploration and play; more often
as they were involved in experimental tasks he had cleverly contrived.
It is probably as an inventor of brilliant experimental paradigms and
riveting demonstrations that Piaget will be best remembered by both the
public and the psychological community. His demonstrations have had
profound effects on work in cognitive developmental psychology. Who,
before Piaget, suspected that infants believe an object remains in its origi-
nallocation even when it has been moved before their eyes to a new one?
Who thought that toddlers cannot appreciate how a collection of objects
looks from a perspective different from their own? Who anticipated kin-
dergarteners' beliefs that the amount of liquid changes when it has been
poured into a vessel of different shape? Even when Piaget's particular
demonstrations have not always stood up in just the way he described
them, further knowledge has invariably been built upon his pioneering
discoveries.
But in Piaget's mind, he was an epistemologist or, as he preferred to
say, a genetic epistemologist-not a child psychologist. He sought to un-
ravel the basic laws of thought: informal tasks with young children were
simply his preferred means of securing data about the nature of knowl-
edge. His principal contribution to psychology, as he saw it, was an un-
raveling of the basic structures of thought that characterize children at
different ages or stages of development, and his suggestion of the mech-
anisms that enable a child to effect a transition to higher stages of devel-
opment-from the sensorimotor stage of infancy to the intuitive stage of
early childhood, or from the concrete operational stage of middle child-
hood to the formal operational stage of adolescence. In positing specific
stages, Piaget relied heavily on logical formalisms; in expounding the
rules of transformation from one stage to another, he relied on biologi-
117
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
118
Psychology: The Wedding of Methods to Substance
of the knowing subject could be adequately controlled. This new psychol-
ogy took many forms in many places, and there were definite swings of
the pendulum between molar and molecular approaches, between an em-
phasis on the determining role of the environment and one on the contri-
butions of the subject to the task at hand. Excesses of introspectionism at
the tum of the century were replaced in tum by excesses of behaviorism
in the early part of the twentieth century. In particular, mentalistic con-
structs were disallowed, and aspects of language and problem solving were
either omitted altogether or treated in much attenuated form.
By the late 1940s, at the time of the Hixon Symposium, it was becom-
ing clear that neither the physiological nor the psychological forms of
behaviorism were viable. Available as alternative models were Gestalt
psychology, as well as the still-isolated efforts to study higher forms of
problem solving by Bartlett, Piaget, and a few other investigators. But it
took the advent of computers (which could themselves exhibit problem-
solving behavior), and the rise of information theory (which provided an
objective basis on which to stipulate components of language or concepts)
to grant legitimacy to cognitive studies. As Ulric Neisser comments, in a
brief sketch of the history of cognitive psychology:
It was because the activities of the computer itself seemed in some ways akin
to cognitive processes. Computers accept information, manipulate symbols, store
items in "memory" and retrieve them again, classify inputs, recognize patterns and
so on. . . . Indeed the assumptions that underlie most contemporary work on
information processing are surprisingly like those of nineteenth century introspec-
tive psychology, though without introspection itself. (1976, pp. 5, 7)
119
II 1 THE CoGNITIVE SciENCES : HisTORICAL PERSPECTIVE
121
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
It was only natural that the modal model became the focus for a generation
of memory textbooks. Textbooks guided by the model characteristically follow the
perceptual input from the sense organs to the brain .... The main problem with
this type of textbook organization is that by the end of the 1970s the "sequential
stage" model [was] modal no longer. (Glass, Holyoak, and Santa 1979, p. iv)
122
Psychology: The Wedding of Methods to Substance
reaches some central processor, where more complex forms of analysis
{including top-down processing) may take place. There has been a correla-
tive tendency to think of an individual's information-processing capacity
as being sharply limited-a certain number of chunks, a certain amount of
sensory information, a limited short-term buffer, and the like; and this
consensus has obtained on both sides of the Atlantic. Both of these ten-
dencies, I should stress, are direct bequests from the Broadbent model of
the human being as a vessel that, like a computer, takes in information and
handles a certain amount of it in a certain amount of time.
Still this general approach to cognition has come increasingly under
critical scrutiny. Some commentators, like Allan Allport (1980) of Ox-
ford, suggest that there are essentially no limitations to the amount of
information that can come in through various sensory channels: on All-
port's account, the primary model of information input should be parallel
(multiple entries at multiple points) rather than serial (one bit after an-
other entering at a single point). Scholars like Richard Shiffrin of Indiana
University have suggested that certain processes can become automatic
and will thereafter exert no significant drain on the organism's ability to
take in or filter novel information. On this view, speed and accuracy
eventually become independent of the number of elements to be proc-
essed (Shiffrin and Schneider 1977). More radical critics, like Benny Sha-
non {1985) of the Hebrew University in Jerusalem, contend that contex-
tual information affects any processing of sensory information: they
question whether any bottom-up or any "outward-inward" model,
which treats "bits of information" apart from their meaning and context
of presentation, can do justice to human cognition. They also challenge
the serial, "one operation at a time" logic of the model. While the stan-
dard linear information-processing mode of thinking has not yet been
abandoned by most researchers, these various chinks in the traditional
picture have led many researchers to rely on parallel, top-down, or con-
textual approaches to the processing of information.
In addition, a closer look at the "modal model" itself pointed up
genuine problems, and the rewards of this research became more elusive.
The more closely the hypothetical stages were examined, the more they
blended into one another; short-term memory could not be readily sepa-
rated from intermediate memory; pre-attentive processes merged with
sensory buffers; processing often did not appear to be strictly serial; the
assumption that specific factors affect specific stages of the sequence
proved difficult to sustain; an individual's expectations might even affect
the early stages of recognition (Glass, Holyoak, and Santa 1979).
Not even the most highly touted experiments withstood close scru-
tiny. While the Sternberg model proved an elegant description of a particu-
123
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
lar experiment, even slight changes in the model affected human perform-
ance in significant ways. For example, if the probability of test items in the
memory set is unequal, reaction times are faster to more probable items;
if an item is repeated in the memory set, reaction times to it are usually
short; if the items in the memory set are drawn from different conceptual
categories, subjects search exhaustively within categories, instead of
searching every category in the memory set. It has even turned out that
some findings can be accounted for equally well using a parallel (as op-
posed to a serial) processing model (Lachman, Lachman, and Butterfield
1979). Because of these disturbing findings, often linked to the content of
items, few strong generalizations of power could be sustained (Glass,
Holyoak, and Santa 1979; Neisser 1976).
But there was an even deeper problem. Even when results held up
reasonably well, there was increasing skepticism about their actual value.
In daily life, one never encounters such meaningless stimuli under such
controlled conditions; meanings, expectations, contextual effects are al-
ways present and are, more often than not, the dominant consideration.
Information-processing psychologists in the Donders-Stemberg-Sperling
tradition developed increasingly elegant models about effects which did
not prove robust when they were changed in various ways; nor did these
models clearly add up to a larger, more comprehensive picture of how
information is processed under real-life situations. Eventually many of
these researchers themselves abandoned this tradition and went on to
other lines of study.
124
Psychology: The Wedding of Methods to Substance
instance, that a subject heard the separate sentences: "The ants ate the
jelly," and, "The ants were in the kitchen." Subjects so stimulated tended
to say that they actually heard the sentence, "The ants in the kitchen ate
the jelly," or even, "The ants in the kitchen ate the jelly which was on the
table" (Bransford and Franks 1971, p. 339).
The Bransford paradigm highlights the inferential and integrative
approaches subjects often exhibit in dealing with language, and thus casts
doubt on the legitimacy of conclusions drawn from the hundreds of ex-
periments in which individuals are asked to remember nonsense syllables
or phrases. Rather than suggesting a passive, rote kind of recall, the Brans-
ford results indicate that subjects are actively and constructively process-
ing information and are inferring meaning rather than recalling sheer
strings of words.
A cluster of further studies from the Bransford group fills out this
perspective on "organizing schemas" (Bransford and Johnson 1972; Brans-
ford and McCarrell1975; Bransford et al. 1979; Johnson, Bransford, and
Solomon 1973). Researchers have found that subjects' abilities to process
a paragraph differ dramatically, depending upon whether they have been
provided with a title or a relevant picture to look at beforehand; the title
or picture creates a set that strongly influences how a sentence is inter-
preted. Other studies reveal that subjects inevitably draw inferences about
the sentences they hear, and answer questions based on those inferences
rather than on the literal contents of the sentences themselves; and that
different cues will help an individual remember a given word, depending
upon which sense of that word has been elicited by the initial sentence in
which it was presented. Such supplementing, interpretive, or inferential
activities tum out to be common if not automatic in the processing of
verbal materials; and, in fact, when meaningful nonverbal materials are
used, be they pictures or songs, the same kinds of "organizing schema" can
be discerned as well.
Once we move beyond the level of sentences and approach para-
graphs, stories, or whole texts, the role of organizing schemas becomes ever
more evident. In an influential line of research, David Rumelhart (1975)
and his colleagues have investigated how individuals remember stories.
Building on some of the ideas in Bartlett's earlier work, and also on
Chomsky's analysis of sentences, the Rumelhart circle has put forth the
notion of a story grammar-an underlying set of assumptions about how the
plot of an ordinary story will unfold. A story-grammatical approach posits
that a story has an actor who seeks various goals, takes steps to achieve
these goals, experiences various reactions in approaching these goals, and
eventually fulfills (or fails to fulfill) them. People bring such expectations
to stories: the stories that fulfill these expectations will prove easy to
125
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
remember; but when stories violate these expectations, people are likely
either to forget the stories or to regularize them (a Ia Bartlett) so that they
become consistent with usual story-grammar assumptions. Moreover,
even when individuals deal with other kinds of experience or text, which
are not storylike but feature a regular sequence of specified events, they
also draw upon organizing scripts or frames: for example, the customary
way in which one changes a tire or visits a restaurant colors one's experi-
ence of, and memory for, new instances of such an incident.
Paralleling trends in other cognitive sciences, this interest in schemas,
scripts, and other inferential and organizing processes has had a major
effect on theory in cognitive psychology. It is not precisely that the belief
in bottom-up or in "sequential serial processing" has been abandoned:
such models still have their uses and adherents. But researchers have come
to appreciate anew that human subjects do not come to tasks as empty
slates: they have expectations and well-structured schemata within which
they approach diverse materials, even including the apparently colorless
and meaningless stimuli of a standard information-processing paradigm.
Thus, an influential alternative approach in cognitive psychology focuses
instead on how the organism, with its structures already prepared for
stimulation, itself manipulates and otherwise reorders the information it
freshly encounters-perhaps distorting the information as it is being as-
similated, perhaps recoding it into more familiar or convenient form once
it has been initially apprehended.
This top-down schematic-based approach stems from earlier efforts in
cognitive psychology: from the "determining tendency" of the Wiirzburg-
ers; from the search for organized structure by Gestalt psychologists; and,
most directly, from Bartlett's story schemas. It also exhibits revealing ties
to Bruner's work on strategies of concept formation. As it happens, the
particular line of work on concept formation undertaken by Bruner, Good-
now, and Austin has rarely been followed up-probably because the view
of concepts as arbitrary clusters of features has been as thoroughly rejected
in cognitive psychology as have Ebbinghaus's nonsense syllables (more on
that in chapter 12). However, interest in the strategies or the organizing
schemes that one brings to a task is a robust legacy of the Bruner enter-
prise.
The adherents of this approach to cognition feel that it is on firm
ecological grounds. Where the users of artificial stimuli and arbitrary tasks
need to prove the relevance of their approaches to the real world, those
who adopt the contextual, top-down approach can point to the fact that,
in everyday experience, such highly interpreted and meaningful stimuli are
the norm.
Even within the relatively conservative area of memory research, there
126
Psychology: The Wedding of Methods to Substance
has been a shift in the perspective of many researchers. Instead of the
Ebbinghaus intoxication with meaningless information, researchers use
meaningful and even highly elaborated, context-rich materials, like stories
(Bower 1976). Attention is now being directed to the "depth" with which
the information is processed. On this view, a subject has the option of
paying attention only to superficial aspects of the stimulus (say, the sounds
of words or the precise syntactic form of phrases) or of assimilating it to
various schemata that have already existed: the more information is en-
veloped in earlier ways of knowing and embedded with rich associations,
the deeper the level of processing; and hence, the more likely that the
information will be firmly encoded and adequately remembered. Whether
the stimulus is processed at a shallow (surface or sensory) level, or at a
more semantically integrated level, depends on the nature of the stimulus,
the time available for processing, and the subject's own motivation, goals,
and knowledge base.
Such a "depth of processing" approach (Craik and Lockhart 1972) has
at least partially replaced sequential models of information processing. On
this view, the nature of incoming information is determined by the opera-
tions performed during its input. Input operations include analysis not
only of sensory aspects but of semantic features as well. Thus, memory
depends on the nature of the code and the type of analysis undertaken, and
not on the properties of particular memory stores. Moreover, one form of
encoding is not necessarily superior to another. Ordinarily it is preferable
to perform deep processing. However, it turns out that a more superficial
mode of processing is recommended if, for example, one is trying to re-
member which words rhyme with an initial list (Lachman, Lachman, and
Butterfield 1979, p. 279). Ultimately, psychology's contemporary memo-
rizer has many options: one can choose to process information at different
depths; and one may even find that, for some purposes, a more shallow
mode of processing is preferable. We have come a long way from the
compulsory stream leading from periphery to long-term memory. Indeed,
in this concern with goal-directedness, diverse approaches to information,
and highly meaningful contents, psychology is re-embracing parts of the
Wiirzburg program.
An interest in peripheral forms of stimulation and in context-free
processing is certainly justifiable, but all too often it seems to leave the
heartland of human psychological processing unscathed. Proceeding in a
molar and top-down fashion, this work on schemas, stories, and depth of
processing strikes many observers (including me) as closer to what psy-
chology ought to focus upon. And yet the question arises whether this
strand of psychology has achieved much that was not evident to our
predecessors or, indeed, to nonpsychologically trained observers. After all,
127
I I I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
Mental Representations
As the molar approach continues to struggle with the molecular ap-
proach, other lines of investigation bear in critical ways on the classical
information-processing model of the 1950s and 1960s. That model was
essentially content-blind: the assumption was that information of any sort is
processed in essentially the same manner. Thus, if short-term memory
holds seven slots, these slots are equally capacious for any kind of informa-
tion, be it verbal, pictorial, or musical, be it dross or gold. By the same
token, if a search takes a certain period of time per item and is exhaustive
rather than self-terminating, these processes obtain irrespective of what is
being searched for and of who is doing the searching.
These assumptions, which were made by all the pioneering cogniti-
vists on both sides of the Atlantic, served as simplifying beginning points.
They also reflected the computer model and information-theoretical ori-
gins of modem cognitive psychology: with such mechanisms, it is rela-
tively easy to set up experiments so that actual contents are irrelevant.
After all, information theory and computers are deliberately constituted to
be content-blind.
Recently, however, these assumptions have been called into question.
One group of scholars, represented by Roger Shepard of Stanford Univer-
128
Psychology: The Wedding of Methods to Substance
sity, has been exploring the operation in individuals of a form of represen-
tation involving visual imagery. Shepard has studied the abilities of in-
dividuals to form mental representations of objects-be they familiar or
unfamiliar shapes-and to answer questions about their physical similari-
ties. In typical studies, subjects are asked to judge whether two presented
geometrical forms are the same when one of them has been rotated, or to
answer questions about other kinds of entities, which can be imaged, such
as the shapes of the fifty United States (Shepard and Chipman 1970).
With his colleague Jacqueline Metzler, Shepard discovered that reac-
tion times preceding decisions about the identity of forms directly reflect
the size of the angle of difference between the orientation of the two forms
(1971). In cases where the angle between the figures is close to zero, a
response is given almost instantaneously: as the angle climbs toward 180
degrees, a subject's reaction times climb in a linear relation to the size of
the angle. The authors interpret such findings as evidence that the subjects
are actually effecting the comparison by mentally rotating one or another
figure at a constant rate. Such mental imagery mimics the trajectory
through which the figures would pass if they were actually in hand and
being rotated before one's eyes.
These experiments, as carried out by Roger Shepard and (in recent
years) by Stephen Kosslyn, are controversial but I shall postpone discus-
sion of such debate to chapter 11. For now, what is germane is Shepard's
belief that cognitive scientists have erred in positing propositional (lan-
guagelike) representations as the lingua franca of cognitive systems. The fact
that computers can-and usually do-transmit information in only one
symbolic form is no reason to assume that human beings do the same.
Shepard sees mental imagery as a human capacity that has evolved over
millions of years to enable the organism to deal with an ever-changing
environment whose consequences it must be able to anticipate. Whatever
the initial focus, such knowledge has been internalized through evolution
so that it is now "pre-wired" in individuals and governs how they appre-
hend objects in space.
Nor are these capacities limited to the visual modality. Shepard's work
with audition suggests that the representation and transformation of musi-
cal sounds reflect the same general principles as does visually presented
information: principles of conservation, symmetry, proximity, and the
like. Thus, Shepard comes to endorse some of the underlying notions of
Gestalt psychology while placing them on more rigorous experimental
footing (Shepard 1981, 1982; see also Jackendoff 1983).
Though growing out of research on information in a single sensory
modality, Shepard's work challenges much of cognitive-psychology main-
stream. If he can make a convincing case for more than one mode, the
129
I I I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
question arises whether there might not be a multiplicity of modes of
mental representation, each tied to a particular content. In the interests of
parsimony, most philosophers and not a few psychologists have voted in
favor of a single form. The positing of another form of mental representa-
tion complicates the picture-for if two representational modes, why not
three, or seven, or three thousand? From Shepard's perspective, however,
it is ill-advised to adjust the findings of human experimentation to the
demands of the digital computer.
Psychology's Contributions
130
Psychology: The Wedding of Methods to Substance
search on visual perception makes little contact with research on story
understanding or musical memory, and often the concepts used to account
for these phenomena are equally remote from one another.
A forthright and ambitious effort to introduce unifying constructs
into cognitive psychology comes from the work over the past decade of
John Anderson (1983). This psychologist, deeply rooted in the practice of
artificial intelligence, has developed the so-called ACT (for "Adaptive
Control of Thought") system which is put forth as a general model of the
"architecture of cognition."
ACT incorporates a process model that describes the flow of informa-
tion within the cognitive system. The central notion is a production sys-
tem (see chapter 6 for further discussion). As soon as a node in a network
is sufficiently activated, a certain action (or production) is carried out: this
construct has been described as a kind of cognitive stimulus-response
bond, because whenever the proper stimulating phenomena are present, an
action is elicited. The system itself includes different kinds of memory-
a working memory (consisting of information with which the system can
currently work), a declarative memory (with propositions in it), and a prOduction
memory (which involves the actions carried out by the system). There are
numerous other mechanisms as well. Encoding processes deposit informa-
tion about the outside world into working memory. Performance processes
convert commands in working memory into behavior. A storage process
creates permanent records in declarative memory; a retrieval process re-
trieves information from declarative memory; and an execution process
deposits the action of matched productions into working memory. A
match process places data in working memory in correspondence with the
conditions of production. Anderson's theoretical work involves specifying
the nature of the knowledge structures that reside in working memory, the
storage process, the retrieval process, and the various factors that activate
productions.
As should be evident from this barrage of mechanisms and concepts,
Anderson's system is complex, as is the evidence he has brought to bear.
The system is also controversial, with some psychologists putting their
faith in the kind of enterprise in which Anderson is engaged, and others
suspecting that it is built on a base of sand. As Keith Holyoak, not an
unsympathetic critic, has noted:
131
I I I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTivE
132
Psychology: The Wedding of Methods to Substance
the ways we humans deal with, say, syntax in language share few funda-
mental properties with our transformations of spatial images or our inter-
pretations of musical expressiveness. None of these possibilities were ade-
quately appreciated by the first generation of psychologist-cognitivists,
and they have been basically ignored or minimized by those who, like
Anderson or Piaget, attempt to produce general cognitive architectures.
If the modularists are on the right track, there is a disturbing possibil-
ity for psychology. Rather than being a single coherent discipline-as its
leading figures have understandably wished-psychology may tum out to
be more of a holding company. In such a vision, there will be separate
studies of language, music, visual processing, and the like, without any
pretense that they fit into one supradiscipline. (In my concluding chapter,
I suggest that this may be a possible fate for cognitive science as a whole.)
Since this outcome would fly in the face of the established wisdom within
the field, it is important to continue to put forth synthetic accounts as
scholars like Anderson are doing.
While current debates pit the generalists against the modularists (see,
for example, Piattelli-Palmarini 1980), it may be that both the modularists
and the central processors have hit upon important truths. The modularists
may be right in thinking that many domains operate by their own laws;
the centralists, in believing in another, synthetic intellectual realm where
modular processes are inadequate, and where horizontal processes prove
necessary. Debate might then come to center on whether any of the modu-
lar realms can be subsumed under some aspect of a central-processing
view.
Discussions of the modular as against the centralist approach are
conducted by those who are generally in sympathy with the current meth-
ods and concepts of cognitive psychology. More severe criticisms of the
enterprise have also been leveled, however. One school deplores the artifi-
ciality of studies in cognitive psychology. In this view, the most convincing
models pertain to tasks that have little clear-cut relation to ongoing human
activity; in contrast, those studies that do pertain to ongoing human activ-
ity take the form of intriguing demonstrations rather than theory-enrich-
ing experiments. In its harshest form, this criticism harbors the suggestion
that the whole information-processing approach has been misguided; that
attempts to model human cognition, in terms of a series of operations upon
mental content, is simply a bad model of the human mind; and that some
radically different (though perhaps not yet articulated) approach will be
needed (Shanon 1984). I shall return to this theme when I consider the
computational paradox.
Such skepticism has even been voiced by some of the leading experts
in psychology. In his 1967 book Cognitive Psychology, Ulric Neisser exuded
133
I I I THE CoGNITivE SciENCES : HISTORICAL PERSPECTIVE
optimism concerning the promise of the field in which he had been work-
ing. But a decade later in Cognition and Reality (1976), he was much more
sober. Berating his colleagues for both their homage to the computer model
and their insistence on artificial laboratory scenarios, he lamented the lack
of an ecologically valid psychology-one that speaks to the kinds of issue
human beings encounter and resolve in everyday life.
As Neisser described it, such a psychology-lacking ecological valid-
ity, indifferent to culture-risks becoming a narrow and uninteresting
specialized field. And, in his view, "the villains of the piece are the me-
chanistic information-processing models, which treat the mind as a fixed-
capacity device for converting discrete and meaningless inputs into con-
scious percepts" (1976, pp. 7, 10). Neisser (1984) calls for a perceptual
psychology, which studies how humans see while moving around and
interacting with objects (Gibson 1979); for concept-formation research,
which features complex objects in the world (Rosch 1973b ); and for a study
of memory which includes autobiographical accounts, eyewitness reports,
and memory of childhood friends (Bahrick, Bahrick, and Wittlinger 1975).
John Morton, one of the leading information-processing psychologists in
Britain, has declared, "Experimental psychology has a disastrous history
with respect to its relevance" (1981, p. 232). And Arnold Glass, Keith
Holyoak, and John Santa, authors of a well-regarded textbook in cognitive
psychology, raise the unsettling possibility that psychology is a "fast race
on a short round track." They explain that this remark is meant to imply:
While Glass and his colleagues do not themselves endorse this bleak out-
look, they correctly note that this view is expressed "with depressing
frequency these days," and, as I have suggested, outside as well as within
the halls of cognitive psychology.
In my own view, there was understandable but probably excessive
excitement about cognitive psychology during the first few years following
its birth (or rebirth). Not surprisingly, then, some of those who were
jubilant at the demise of behaviorism have been less than ecstatic at the
results of the last (or first!) twenty-five years. There is certainly nothing
wrong with experiments conducted under artificial conditions per se (see
134
Psychology: The Wedding of Methods to Substance
the example of physics!), but there is something bothersome when findings
prove fragile once the experimental conditions are altered slightly, applied
to more complex phenomena, or moved one step outside the laboratory.
Part of the dissatisfaction at the present time comes, paradoxically,
from the perfection of experimental methods within psychology. One of
the genuine contributions of behaviorism in psychology was the refine-
ment of experimental methodology; but in this case, the refinements may
have reached the point of diminishing returns. As I read the literature, too
much emphasis has been placed on having an experimental procedure
without any perceptible flaws or ambiguities; and all too often this empha-
sis takes place at the expense of considering what is an interesting or
important problem. By the same token, once an interesting demonstration
has been introduced into the psychological literature, dozens of other
experimenters devote their attention-and their methodological zeal-to-
ward finding the vulnerabilities in the experiment. And ultimately, nearly
all these experiments get shown up (and ultimately abandoned) for their
limitations. These trends combine to produce a science that has progressed
rather less than it should have, and that still consists more of a set of
impressive but isolated findings than of a truly cumulative discipline.
It hardly needs saying that it has been easier to achieve methodolog-
ical perfection in those studies that deal with molecular phenomena and
proceed in a bottom-up fashion. As I have noted, however, there recently
has been growing dissatisfaction with experiments in this tradition: and
there has been a correlative attraction for psychological investigations that
deal with more molar phenomena (like the understanding of a story or the
transformation of a mental map) and that do so in a top-down fashion
(taking into account schemas, frames, and the surrounding context). The
pendulum has definitely been swinging in a molar direction once again.
These two situations-the intoxication with method and the trend
toward molarity-provide psychology with an invaluable opportunity.
Put simply, it is time for psychology to wed its indubitable methodological
sophistication to a concern for problems that are more molar, less artificial,
more representative of real-life situations, more substantive. Just how to
accomplish this is hardly an easy question to answer-though I have tried
to indicate some of the lines of investigation that I personally find most
promising. In this context, I strongly believe that psychologists can benefit
from studying the example of artificial intelligence. In spite of (or perhaps
because of) the presumed rigor of the computer, researchers in the A.l.
tradition have been much more ready to deal with molar topics (like
problem solving or story understanding) and to do so in a top-down
fashion. The partnership that is developing between the two fields-and
that will occupy me more fully in the next chapter-may serve as the best
135
II I THE CoGNITivE SCIENCES : HISTORICAL PERSPECTIVE
136
Psychology: The Wedding of Methods to Substance
While I expect that cognitive psychology will simply, and advanta-
geously, be absorbed into an issue-oriented cognitive science, I do not
mean to indicate that psychology as a whole will disappear. (Far be it for
me to predict-or wish-the end of my own discipline!) There will always
be many practical needs for psychologists in the clinic, in the school, and
in industry. Even on the level of theory, some pockets of psychology will
endure. I do not anticipate that topics like human personality or motiva-
tion will be absorbed by other disciplines: neither neuroscience nor artifi-
cial intelligence can gobble them up. By the same token, psychology will
continue to be indispensable in illuminating the differences among human
beings: between normal and exceptional individuals, between individuals
with different kinds of disease, or individuals from different social or
cultural settings. The practice of individual or differential psychology
should continue and may well yield interesting scientific principles. By the
same token, psychology can play an equally valuable role in effecting other
kinds of comparison-between one sensory modality and another, be-
tween one symbol system and another, between one form of representa-
tion and the other-whether or not, in the end, the modularity view is
judged as more powerful than the central-processing view.
Psychology has been one of the great disciplinary success stories of
this century, and it seems inopportune to be predicting its disappearance
into a larger discipline. In fact, however, it is precisely the innovations in
psychology of the past few decades which have given rise to a larger
cognitive revolution, within which psychology has clearly been a central
discipline. In no ways should psychology be considered simply a holding
operation until neurology or sociology or anthropology comes along: as
much as any scholars, psychologists have successfully made the case for
the centrality of the human mind and mental representation within the
current scientific milieu. The ultimate merger of psychology with artificial
intelligence and other disciplines into a larger cognitive science will be a
tribute to its success and its significance.
Perhaps events of the past century have demonstrated that psychol-
ogy is a difficult science to bring to completion, at least without the aid of
other disciplines. But even more clearly, we have come a long way since
Kant declared, less than two centuries ago, that a psychological science was
impossible.
137
6
Artificial Intelligence:
The Expert Tool
I am prepared to go so far as to say that within a few years,
if there remain any philosophers who are not familiar with
some of the main developments in artificial intelligence, it
will be fair to accuse them of professional incompetence,
and that to teach courses in philosophy of mind, episte-
mology . . . without discussing . . . aspects of artificial
intelligence will be as irresponsible as giving a course in
physics which includes no quantum theory.
-AARON SLOMAN
138
Artificial Intelligence: The Expert Tool
the founder and first director of the A.I. labs at both the Massachusetts
Institute of Technology (1957) and Stanford University (1963). McCarthy
was the major organizer of the institute and the coiner (according to most
accounts) of the term artificial intelligence. The remaining three leading figures
were Marvin Minsky, then a Junior Fellow in mathematics and neurology
at Harvard and eventually the director of the Artificial Intelligence Labora-
tory at M.I.T.; and Herbert Simon and Allen Newell, then at the Rand
Corporation in Santa Monica and also at Carnegie Institute of Technology
(now Carnegie-Mellon University) in Pittsburgh, where they have re-
mained until this day.
The summer at Dartmouth gave these and other scholars a chance to
exchange views and to arrange to collaborate on future work. Various
authorities during the 1940s and early 1950s had expressed the belief that
computers should be able to carry out processes resembling human think-
ing, and it was the job of the present assemblage to put these promises to
the test. Alex Bernstein, then a programmer for International Business
Machines in New York City, talked about the chess-playing program on
which he was working. Arthur Samuel, also of the I.B.M. Corporation,
discussed a program that played checkers. Newell and Simon described a
program that they had devised to solve theorems in logic. Nathan Roches-
ter of I.B.M. in Poughkeepsie described work on programming a model of
neural networks, while Marvin Minsky discussed the use of computers to
prove Euclidean theorems.
The meeting at Dartmouth did not fulfill everyone's expectations:
there was more competition and less free exchange among the scholars
than the planners had wished. Nonetheless, the summer institute is con-
sidered pivotal in the history of the cognitive sciences, in general, and in
the field of artificial intelligence, in particular. The reason is, I think, chiefly
symbolic. The previous decade had seen the brilliant ideas of an older
generation-Norbert Wiener, John von Neumann, Warren McCulloch,
Alan Turing-all point toward the development of electronic computers
which could carry out functions normally associated with the human
brain. This senior group had anticipated developments but were uncertain
whether they themselves would have the opportunity to explore the
promised land.
At Dartmouth, members of a younger generation, who had grown
up in an atmosphere seeded with these seminal ideas, were now ready
(and in some cases, beyond mere readiness) to devise the machines and
write the programs that could do what von Neumann and Wiener had
speculated about. These younger scholars were attracted by powerful (if
still vague and poorly understood) notions: data being processed by a
program and then becoming part of the program in itself; the use of
139
II I THE COGNITIVE SCIENCES : HISTORICAL PERSPECTIVE
Since the Dartmouth meeting, artificial intelligence has had a brief but
stormy history. Part of the storm swirls around definitions. Nearly all
authorities agree that artificial intelligence seeks to produce, on a com-
puter, a pattern of output that would be considered intelligent if dis-
played by human beings. Most authorities see the computer program as
a test of a particular theory of how a cognitive process might work. But
thereafter consensus falters. Some definitions stress the devising of pro-
grams; others focus on programming languages; others encompass the
mechanical hardware and the human conceptual component, as well as
the software. Some practitioners want to simulate human thought pro-
cesses exactly, while others are content with any program that leads to
intelligent consequences.
Authorities also disagree on how literally to take the thinking meta-
phor. Some researchers take what has come to be termed the "weak view,"
where the devising of "intelligent" programs is simply a means of testing
theories about how human beings might carry out cognitive operations.
Others, however, put forth much more forceful claims about their field.
According to the view of "strong AI," as phrased by philosopher John
Searle, "the appropriately programmed computer really is a mind, in the
sense that computers given the right programs can be literally said to
understand and have other cognitive states. In strong AI, because the
programmed computer has cognitive states, the programs are not merely
140
Artificial lnfelligence: The Expert Tool
tools that enable us to test psychological explanations; rather the programs
are themselves the explanations" (1980, p. 417). I shall consider the merits
of weak and strong A.I. at the conclusion of this chapter.
But while the tension between weak and strong claims is one of the
most momentous debates, it is by no means the only one. As Robert
Wilensky, a leading artificial intelligence researcher, recently commented,
"Artificial intelligence is a field renowned for its lack of consensus on
fundamental issues" (1983, p. xii). Indeed, in a recent capsule history of
artificial intelligence, Allen Newell (1983) was able to single out no fewer
than three dozen issues that have at times divided the field. While some
of these are quite technical, and others of only transient interest, two of
them seem to me of particular note. The first is the tension between
"generalists" and "experts"-a tension recalling the dialectic between the
modular and the central-processing perspectives in contemporary psy-
chology. Generalists believe in overarching programs (or families of pro-
grams) that can be applied to most any manner of problem: experts place
their faith in programs that contain much detailed knowledge about a
specific domain but prove relatively restricted in their applicability. A
second tension has to do with the scientific status of the field. While
some of the founders were prepared to make strong claims for scientific
importance (and, indeed, see A.I. as replacing epistemological pursuits),
more skeptical commentators have wondered whether artificial intelli-
gence deserves to be considered a scientific discipline at all. From their
point of view, A.I. is simply a form of applied engineering-even gim-
mickry-with no real standing as a theoretically based scientific disci-
pline. To be sure, similar skeptical challenges have been leveled at other
of the cognitive sciences; but perhaps because of the dramatic promise of
a "thinking machine," the battles about the scientific status of A.I. have
been particularly vehement.
In this chapter, I shall observe the wide swings of mood and the
diverse viewpoints that have characterized the leading practitioners and
commentators during the first three decades of A.I. It is not possible, of
course, to touch on every strand of artificial intelligence; for example,
except incidentally, I shall not discuss work on robots, retrieving informa-
tion from data bases, planning optimal combinations or optimal schedules,
simulating organizational activities, or writing programs that write pro-
grams, even though each of these areas is becoming part of standard
reviews of artificial intelligence (Nilsson 1980; Waldrop 1984a, 1984c;
Winston 1977). But I will touch on those lines of work that seem to me
to be most relevant to human psychology and, at the conclusion of this
chapter, attempt to situate the field of A.I. within the broader framework
of cognitive science.
141
II I THE CoGNITIVE SciENcEs : HISTORICAL PERSPECTIVE
142
Artificial Intelligence: The Expert Tool
there were operation cards, which directed the operations to be per-
formed, and variable cards, which determined the particular variables on
which the operations were to be performed (Dorf 1974). Any arithmetic
problem could be set; and, provided that the proper cranks were turned,
the right answer should issue forth.
While Babbage was attempting to implement his ambitious mechani-
cal aspirations, another British mathematician, George Boole of Queens
College Cork, was involved in a different but equally momentous under-
taking: that is, to figure out the basic laws of thought and to found them
on principles of logic. In order to eliminate the ambiguities of natural
language (which had dominated logic since the time Aristotle had studied
the syllogism), Boole used a set of arbitrary symbols (a, b, x, y, and so on)
to stand for components of thought. As he put it, "a successful attempt to
express logical propositions by symbols, the laws of whose combinations
should be founded upon the laws of the mental processes which they
represent, would, so far, be a step toward the philosophical language"
(quoted in Hilton 1963, p. 163). These symbolic elements could be com-
bined or dissociated through operations like adding, subtracting, or multi-
plying, so as to form new expressions, or new conceptions, involving the
same elements. These procedures amounted to a kind of "mental algebra,"
where reasoning could be carried out in abstract positive or negative terms,
unsullied by the particular associations tied to specific contents. And these
operations were termed by Boole the "laws of thought." Most important
for the future, Boole observed that his logic was a two-valued or true-false
system. Any logical expression, no matter how complex, could be ex-
pressed either as 1 (standing for "all," or "true"), or as 0 (standing for
"nothing," or "false"). The idea that all human reason could be reduced
to a series of yes or no discussions was to prove central for the philosophy
and the science of the twentieth century.
The significance of Boole's work was finally appreciated a half-cen-
tury later by Alfred North Whitehead and Bertrand Russell when they
produced their Principia Mafhemafica (1910-13). The goal of this work, as I
have noted, was to demonstrate that the roots of mathematics lie in the
basic laws of logic. Whitehead and Russell relied heavily on the formalism
pioneered by Boole. Russell went so far as to declare in his ascetic manner,
"Pure mathematics was discovered by Boole in a work he called 'The Laws
of Thought'" (quoted in Halacy 1962, p. 106).
The cluster of ideas represented by Babbage's calculating machines,
Boole's laws of thought, and Whitehead and Russell's decisive demonstra-
tions were eventually to be integrated by scholars in the 1930s and 1940s.
Their work culminated in the first computers and, eventually, the first
programs that can be said to display intelligence.
143
II I THE COGNITIVE SCIENCES : HISTORICAL PERSPECTIVE
144
Artificial Intelligence: The Expert Tool
plinary field which investigated feedback mechanisms in organic matter
and in automata.
Finally, there was John von Neumann, in touch with all of these veins
of thought, and with perhaps the most sustained interest in the theory of
computing machines. He is generally credited with developing the idea of
a stored program, where the operations of the computer can be directed or
controlled by means of a program or a set of instructions, housed in the
computer's internal memory. (It was thus no longer necessary to reprogram
for each new task.) He demonstrated how binary logic and arithmetic
could work together in forming stored programs. One can encode instruc-
tions to the machine in the same language as that used for the data it
processes, and thus mix instructions and data in the program and store
both in the computers. These conceptual breakthroughs opened the way
for adjuncts to programming-such as assemblers, which can cull together
subroutines into the main program, and compilers, which can translate from
one language (usually a high-level programming language which is conve-
nient to use) to a more basic language, reflected in the actual electrome-
chanical operations of the computers. Finally, von Neumann pursued with
special vigor the analogies (and disanalogies) between the brain and com-
puting machines.
It is not entirely clear whether von Neumann appreciated the potential
of programs to attack and solve problems of intellectual depth. Yet he was
certainly aware of the nexus of issues in this area and if he had not died
of cancer while still relatively young, he might well have become the major
figure in the history of artificial intelligence. But this role was shared by
four scholars at the Dartmouth meeting in 1956: Herbert Simon and Allen
Newell, Marvin Minsky, and John McCarthy.
145
I I I THE CoGNITIVE SciENCES : HisTORICAL PERSPECTIVE
146
Artificial Intelligence: The Expert Tool
The demonstration that the Logic Theorist could prove theorems was
itself remarkable. It actually succeeded in proving thirty-eight of the first
fifty-two theorems in chapter 2 of the Principia. About half of the proofs
were accomplished in less than a minute each; most of the remainder took
from one to five minutes; a few took fifteen to forty-five minutes; there
was a strong relation between the number of items in the logical expression
and the length of the proofs. It turned out that one proof was more elegant
than Whitehead and Russell's attempt of fifty years before-as Simon
informed Bertrand Russell, who was delighted by this ironic twist. How-
ever, the journal of Symbolic Logic declined to publish an article co-authored
by the Logic Theorist in which this proof was reported (McCorduck 1979,
p. 142).
It was still possible that this demonstration-although intriguing to
engineers or logicians-could fall outside the purview of scientists inter-
ested in the operation of the human mind. But Newell, Simon, and their
colleague Cliff Shaw stressed that they were demonstrating not merely
thinking of a generic sort but, rather, thinking of the lcind in which humans
engage. After all, Logic Theorist could in principle have worked by brute
force (like the proverbial monkey at the typewriter); but in that case, it
would have taken hundreds or even thousands of years to carry out what
it actually achieved in a few minutes. Instead, however, LT worked by
procedures that, according to the Newell team, were analogous to those
used by human problem solvers. Among the methods used by LT are
substitution of one kind of expression for another; a detachment method,
where the program works backward from something that has already been
proved to something that needs to be proved; and a syllogistic form of
reasoning, where if "a implies b" is true, and "b implies c" is true, then
'a implies c" is also true.
In a further effort to underscore parallels between human and machine
problem solving, Newell and Simon performed various experiments with
their program. They showed that, if they removed the record of previous
theorems (on which solutions to new theorems were constructed), the
Logic Theorist could not solve problems it had previously handled in ten
seconds. This was perhaps the first attempt ever to perform an experiment
with a computer to see if it "responds" in the way that human beings do.
Trying to portray their demonstration in the proper light, the Newell
team also stressed the resemblance between human and machine problem
solving. They based this claim on some protocols they had gathered on
human subjects engaged in the same tasks. In both humans and computing
machines, the team found certain staples of human problem solving. For
example, they reported certain moments of apparent insight as well as a
reliance on an executive process that coordinates the elementary opera-
147
I I I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
We do not believe that this functional equivalence between brains and com-
puters implies any structural equivalence at a more minute anatomic level (for
example, equivalence of neurons with circuits). Discovering what neural mech-
anisms realize these information processing functions in the human brain is a task
for another level of theory construction. Our theory is a theory of the information
processes involved in problem-solving and not a theory of neural or electronic
mechanisms for information processing. (Newell, Shaw, and Simon 1964, p. 352)
148
ArH/icial Intelligence: The Expert Tool
The methods used by the General Problem Solver can be readily
described. In means-ends analysis, one first states the desired form of the
solution of a problem, and then compares one's present place in the process
of problem solution with the final goal desired. If these two instances
coincide, then the problem has been solved. If not, the solver (human or
mechanical) clarifies the difference and searches for methods to reduce the
difference between where one is and where one wants to go.
The art in the General Problem Solver lies in the methods of reducing
this distance. A table is set up that associates the system's goals with
operators that may be of use in achieving them. Once the difference has
been computed between the present situation and the goat the system
then selects an operator associated with that difference and tests whether
the operator is applicable to the current situation. If it can be applied, and
if it produces a result that is closer to the desired end state, it is repeated
again. If it proves inapplicable, then the system generates a subgoal, whose
aim is to reduce the difference between the current situation and the
situation where the operator can be applied. This procedure is simply
repeated until the goal is achieved or it has been demonstrated that it
cannot be achieved with the information given, or with the operators
available in the program.
The General Problem Solver also exhibited other features designed to
facilitate problem solving. It was possible to decompose the program into
subproblems which could be tackled one at a time. It was possible to ignore
some of the complicating factors in a situation in order to arrive at a plan
of attack. It was possible to omit certain details of a problem as well. For
example, in solving a problem in the propositional calculus, the machine
can decide to ignore differences among the logical connectives and the
order of symbols, only taking into account what the symbols are and how
they have been grouped.
While the General Problem Solver was eventually abandoned because
its generality was not as great as its creators had wished, and because the
field of artificial intelligence moved in different directions, the program can
be regarded as the first to simulate a spectrum of human symbolic behav-
ior. GPS also occupied a major role in Simon and Newell's thinking about
the enterprise in which they were engaged. As they conceived it, all intelli-
gence involves the use and manipulation of various symbol systems, such
as those featured in mathematics or logic. In the past, such manipulation
had been done only by the human being within the confines of his own
head, or with paper and pencil, but, with the advent of the digital com-
puter, symbol manipulation has become the province of electronic machin-
ery as well. On the Newell-Simon account, the computer is a physical
symbol system like the human brain and exhibits many of the same prop-
149
II I THE CoGNITIVE SciENCES : HisToRICAL PERSPECTIVE
erties as the human being whose behavior it has now been programmed
to simulate.
Just as the cell doctrine has proved central in biology, and germ theory
is pivotal in the area of disease, so the concept of physical symbol system is
deemed by Simon and Newell and their colleagues at Carnegie-Mellon as
the core doctrine in the area of computer science. Proceeding in Boolean
spirit, the job of the theorist is to identify that set of processes which
operates on symbolic expressions in order to produce other expressions
that create, modify, reproduce, and/or transform symbolic structures. A
physical symbol system is necessary and sufficient to carry out intelligent
actions; and, conversely, any system that exhibits general intelligence will
prove upon analysis to be a physical symbol system. Such a system con-
sists of a control, a memory, a set of operations, and input and output: its
input consists of objects in certain locations; its processes are operations
upon the input; its output is the modification or re-creation of objects in
certain locations.
A key notion in the Newell-Simon scheme is the production system,
in which an operation will be carried out if a certain specific condition is
met. Programs consist of long sequences of such production systems oper-
ating on the data base. As described by the theorists, the production
system is kind of a computational stimulus-response link; so long as the
stimuli (or conditions) are appropriate, the response (or production) will
be executed. In the course of developing the General Problem Solver,
Simon and Newell had propounded a perspective on artificial intelligence,
a theory of thinking, and an agenda for future research.
The vision of Newell and Simon was formidable. From their perspec-
tive, the profound similarities between the human mind engaged in solving
a problem, and the computer programmed to solve the same problem, far
overrode differences in hardware (an electronic machine versus a parcel of
neural tissue). Both are simply systems that process information over time,
proceeding in a more or less logical order. Moreover, to the extent that the
steps noted by an introspecting individual paralleled the lines of a com-
puter program, one was no longer simply engaging in weak A.I.: it made
sense to think of this man-made physical symbol system as actually
engaging in problem solving.
Critics of the Newell-Simon effort brought up a number of issues.
First of all, all the information in the computer program had been placed
inside the program by humans: thus, to put it colloquially, the problem
solver was only doing what it was programmed to do. For instance, it
was Newell and Simon who structured the problems given to the pro-
gram and, in some cases, determined the order in which they were pre-
sented. Use of terms like insight was but a misleading metaphor. To
150
Artificial Intelligence: The Expert Tool
Newell and Simon, this criticism appears anachronistic: so long as the
program was not simply engaging in rote repetition of steps but actually
used rules to solve problems to which it had not previously been ex-
posed, its behavior was as "intelligent" as that of a human being. Simply
old-fashioned habits of thought induced critics to withhold the term
intelligent. Indeed, it was necessary to think of humans as being pro-
grammed with rules, just like computers. Not all scholars were convinced
by this "democratic response."
Another line of criticism centered on certain differences between
human beings and computer programs. For example, human beings can
improvise shortcuts or heuristics, whereas computers will repeat the same
processes unless they can be programmed to learn from earlier efforts.
Recognizing this limitation in the General Problem Solver, Newell and
Simon set out to devise programs capable of learning.
A final line of criticism involves the kinds of problem posed to the
General Problem Solver. Despite its ambitious name, the problems were all
puzzles or logical challenges: these lend themselves to expression in sym-
bolic forms, which in turn can be operated upon. This restriction to
"closed" questions was essential since GPS could tackle only mathemati-
cal-logical problems. Clearly, many problems confronted by humans (such
as finding one's way about a forest or learning to master a dance) are not
readily expressed in symbolic logic. Here we encounter a revealing instance
of how notions of "thinking" or "problem solving" may be artificially
constrained by the programs that currently exist.
Some of these criticisms pertain to other efforts in artificial intelligence
as well, but each separate line of inquiry deserves consideration on its own
merits. Let me therefore turn more briefly to what other principal inves-
tigators were doing in the first decade or so following the Dartmouth
conference.
151
I I I THE CoGNITIVE SciENCES : HisToRICAL PERSPECTIVE
The gas consumption of my car is 15 miles per gallon. The distance between
Boston and New York is 250 miles. What is the number of gallons of gas used on
a trip between New York and Boston? (Bernstein 1981, p. 113)
From the mathematical word "per" in that first sentence's "miles per gallon,"
it can tell that the number 15 would be obtained by dividing a certain number
x of miles, by some other number, y, of gallons. Other than that, it hasn't the
slightest idea what miles or gallons are, or, for that matter, what cars are. The
second sentence appears to say that something else equals two hundred and fifty
miles-hence the phrase "the distance between" is a good candidate to be x. The
152
Artificial Intelligence: The Expert Tool
third sentence asks something about a number of gallons-so that phrase of "gas
to be used on a trip" is a candidate to bey. So it proposes one equation: x = 250,
and another equation, xly = 15. Then the mathematical part of the program can
easily find that y = 250/15. (Bernstein 1981, p. 113)
153
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTivE
to the few numbers each recognizes, and make the best guess about the
mathematical operations that should be carried out on those numbers.
While the problems selected by Minsky and students span a wider gamut
than those tackled by GPS, they are always reformulated as symbolic
expressions of a canonical sort. Nonetheless, it must be stressed that, as
with the Newell-Simon efforts, the programs worked. The kinds of exer-
cise expressed in ordinary language, over which schoolchildren have
struggled for generations, could be solved by a mechanical process.
154
Artificial Intelligence: The Expert Tool
Patrick Hayes, then of Edinburgh University, McCarthy wrote a seminal
article in 1969, in which he argued for the use of a formal predicate calculus
substrate embedded within a system designed for understanding language
(McCarthy and Hayes 1969). He called for the formalization of concepts
like causality, ability, and knowledge. If such an approach were to be adopted,
it would prove possible to use theorem-proving techniques that are not
dependent on the details of particular domains. McCarthy's system was
based upon a faith in the consistency of a belief system and upon a view
that all knowledge is (or can be) thought of in purely logical terms. As we
shall see, these assumptions have had relatively few proponents in recent
years. Still, McCarthy has adhered to the general program. He has been
designing a nonconventional modification of standard logic in order to
model common-sense reasoning, and his former associate Hayes (1982) has
been trying to formulate in logical terms the thinking processes of "naive
physics"-the physics of the man in the street. McCarthy stands as an
extreme proponent of one point of view in artificial intelligence-a man
who holds high standards for the field and is less willing than others to
bend to the practical or to the particular.
While mainstream scholars in the 1960s were either pursuing logical prob-
lem solving or theorizing about the proper course of artificial intelligence,
certain other work undertaken during that era held considerable implica-
tions for the future. One such line was carried out by Edward Feigenbaum,
an early student of Simon's. In collaboration with Joshua Lederberg, a
Nobel Laureate in genetics at Stanford, Feigenbaum decided to analyze
data from a mass spectrograph (Feigenbaum, Buchanan, and Lederberg
1971). The result was a program named DENDRAL which was designed
to figure out (on the basis of a vast amount of data from mass spectro-
graphs) which organic compound was being analyzed. After processing the
relevant data, the program formulated hypotheses about the compound's
molecular structure and then tested those hypotheses by way of further
predictions. The ultimate output was a list of possible molecular com-
pounds ranked in terms of decreasing plausibility. DENDRAL's perform-
ance compared favorably with that of expert chemists (Boden 1977).
As a program, DENDRAL departed from Simon and Newell's classical
model in two ways. First of all, DENDRAL made use of considerable stored
knowledge about chemicals. It was predicated on the assumption that
experts have a great deal of knowledge and as much of it as possible ought
to be possessed by any program that purports to behave intelligently.
155
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
Second, if seemingly at odds with this assumption, was the decision not
to try to emulate the ways in which human beings would solve the prob-
lem. Feigenbaum was explicitly interested in top-flight performance, and
not particularly intent on mimicking the way that a human being would
operate when confronted with the same closetful of information. Here we
see an early sign of the sharp cleavage that was to ensue between simula-
tion of human behavior and cultivation of intelligent performances by
whatever means.
Two other investigative lines in the 1960s attempted to enter into a
seemingly elusive area of human performance: meaningful conversation
between individuals. In view of Turing's famous challenge (see pages
17-18), this was a tempting goal to pursue. Kenneth Colby, a psychiatrist
then working at Stanford, and Joseph Weizenbaum of M.I.T. both wrote
programs that modeled two-person dialogues. In Colby's case, he was
interested in psychiatric disorders and so produced a program whose an-
swers might be given by a paranoid personality with a particular obsession
about the Mafia:
156
Artificial Intelligence: The Expert Tool
DO YOU THINK COMING HERE WILL HELP YOU NOT TO BE UN-
HAPPY?
I need some help, that much seems certain.
WHAT WOULD IT MEAN TO YOU IF YOU GOT SOME HELP?
Perhaps I could learn to get along with my mother.
TELL ME MORE ABOUT YOUR FAMILY.
(Boden 1977, pp. 106-7; see also Weizenbaum 1966)
157
I I I THE CoGNITIVE SciENcEs : HISTORICAL PERSPECTIVE
Find a block which is taller than one you are holding and put it into the box.
OK
What does the box contain?
THE BLUE PYRAMID AND THE BLUE BOX.
158
Artificial Intelligence: The Expert Tool
What is the pyramid supported by?
THE BOX
Will you please stack up both of the red blocks and either a green cube or a
pyramid?
OK
Which cube is sitting on the table?
THE LARGE GREEN ONE WHICH SUPPORTS THE RED PYRAMID.
Is there a large block behind a pyramid?
YES, THREE OF THEM: A LARGE RED ONE, A LARGE GREEN CUBE
AND THE BLUE ONE.
Thank you.
YOU'RE WELCOME.
(Winograd 1972, quoted in Boden 1977, pp. 115-22)
159
II ·I THE CoGNITIVE SciENCES : HisTORICAL PERSPECTIVE
mean task) and the temporal sequence of all actions performed. In addition
to these experts which can flexibly interact with one another and share
information, the program also has belief sys,tems, knowledge about prob-
lem solving (such as the mechanisms of deduction), and specialists that
detect whether an utterance is question, command, or comment.
SHRDLU was, for its time, a stunning demonstration of apparent
language production and understanding capacities. Yet, according to Dan-
iel Dennett (1978), one of its chief contributions lay in another sphere.
Specifically, the SHRDLU experiment explored some of the extensive de-
mands imposed on any system that undertakes to follow directions, plan
changes in the world, and then keep track of them. While the ways in
which these tasks are done may not simulate the way that a real person
would work in a real block world, the procedures devised by Winograd
were ingenious and at least suggested the kinds of problem that would
have to be confronted by any system seeking truly to understand, rather
than simply to mimic understanding.
Certain limitations should be noted, however. SHRDLU does not have
adequate semantic information to appreciate the differences in meaning
between function words like and, the, and but. More tellingly, SHRDLU
lacks any ability to leam to perform better. Its knowledge suffices for it to
know why it does what it does but not enough for it to remember what
has gone wrong when a failure occurs, or to learn from error to make more
appropriate responses in the future. A later block-world program, designed
by Gerald Sussman (1975) and called HACKER, showed that such learning
was, in fact, possible.
Winograd's program came into being at a crucial moment in the his-
tory of artificial intelligence. There were furious debates going on about
that time, both within the artificial intelligence community and between
it and its severest critics. In many ways, the Winograd program spoke to
these debates, even if it did not singlehandedly resolve them.
Pivotal Issues
160
Artificial Intelligence: The Expert Tool
mater, to deliver a talk to an audience that included Newell, Simon, and
other leading cognitive scientists. Feigenbaum threw out a challenge to his
former teachers: "You people are working on toy problems. Chess and
logic are toy problems. If you solve them, you'll have solved toy problems.
And that's all you'll have done. Get out into the real world and solve
real-world problems" (Feigenbaum and McCorduck 1983, p. 63). Feigen-
baum had brought along information about DENDRAL, his expert system
which incorporated massive amounts of specific knowledge about organic
compounds. While DENDRAL was achieving impressive successes, it ini-
tially met a skeptical audience. As Joel Moses of M.I.T. has commented:
The word you look for and you hardly ever see in the early AI literature is
the word knowledge. They didn't believe you have to know anything, you could
always rework it all .... In fact 1967 is the turning point in my mind when there
was enough feeling that the old ideas of general principles had to go. . . . I came
up with an argument for what I called the primacy of expertise, and at the time
I called the other guys the generalists. (Quoted in McCorduck 1979, pp. 228-29)
He called my position the big-switch theory, the idea being that you have all
these experts working for you and when you have a problem, you decide which
expert to call in to solve the problem. That's not AI, you see.... I think what finally
broke Newell's position was Winograd. (Quoted in McCorduck 1979, p. 229)
And, indeed, shortly after Winograd's program had been completed, even
the leaders of the old guard, like Marvin Minsky and Allan Newell, be-
came convinced of the limitations of generalist programs and of the need
for systems possessing considerable specialized or expert knowledge. Yet
analogous issues were to return in other guises: for example, the question
whether there should be a general language for all artificial intelligence
programs or many specifically crafted tongues.
161
I I I THE CoGNITivE SciENCES : HISTORICAL PERSPECTIVE
once and can then be tapped at will. Declarativists felt that intelligence
rests on a highly general set of procedures which can be used widely,
coupled with a set of specific facts useful for describing particular
knowledge domains; they were not convinced that knowledge of a sub-
ject matter is necessarily bound up with the procedures entailed in its use
in one or another context.
In a manner reminiscent of the functionalists in psychology, the
proceduralists felt that human intelligence is best thought of as a set of
activities that individuals know how to do; whatever knowledge is neces-
sary can be embedded in the actual procedures for accomplishing things.
Many things that we do know how to do are best viewed as procedures;
and, indeed, it is difficult to describe them formally in a declarative way.
For example, if one wants to build a robot to manipulate a simple world,
one does it most naturally by describing these manipulations as procedural
programs. Procedural representation has the additional advantages of al-
lowing ready use of higher-order (second-level) control information or
knowledge of where one routine in a program has to be triggered by
another, and being applicable across several domains (Boden 1977; Cohen
1977; Newell 1983; Winograd 1975).
The debate about procedural versus declarative representation has
become somewhat muted in the last several years. There is now increasing
recognition that not all computing functions lend themselves better to one
mode of representation than to another; and that, in fact, some problems
are better handled by one approach, other problems by its rival. Initially
Winograd had been a chief proponent of procedural systems; but he him-
self wrote a paper describing the advantages of these two different modes
and, more recently, has collaborated with Daniel Bobrow in devising sev-
eral Knowledge Representational Languages which incorporate both
procedural and declarative components (Bobrow and Winograd 1977).
162
Artificial Intelligence: 77ze Expert Tool
and that a dangerous confusion was taking place regarding what was
properly the realm of humans and what ought to be ceded to machines.
People are entities that are wholly different from machines; such uniquely
human experiences as love and morality must remain sacrosanct.
In an even more wide-ranging attack on artificial intelligence, Hubert
Dreyfus, a phenomenologically oriented philosopher at the University of
California in Berkeley, had published in 1972 a critical book called What
Computers Can't Do: A Critique of Artificial Reason. Dreyfus made much in this
book of the fundamental differences between human beings and comput-
ers: he claimed that, unlike computers, human beings have a fringe con-
sciousness; a tolerance for ambiguity; a body that organizes and unifies
one's experience of objects and subjective impressions; the potential for
boredom, fatigue, or loss of drive; and clear purposes and needs that
organize one's situation.
According to the tale Dreyfus spins, after an initial run of apparent
successes, artificial intelligence had bogged down because it had no way
of coming to grips with these fundamental differences between human
beings and machines. It was wedded to the notion that all human behavior
-including all intelligence-can be formally described by logical rules.
But human life is only as orderly as necessary; it is never completely
rule-governed: life is what humans make it and nothing else. Since a
computer is not involved ("engaged") in a situation, since it has no needs
or wants or purposes, it must treat all facts as equally relevant at all times:
it cannot make the kinds of discrimination and evaluation that are the stuff
of human life and make it meaningful.
Weizenbaum's book was greeted critically, but with a modicum of
respect, by his colleagues in artificial intelligence; on the other hand, hardly
any computer scientist had a good word to say about Dreyfus's harsh
verdict. Clashes of personality and even charges of intellectual incompe-
tence came to dominate the debate. Regrettably, there was little serious
discussion between critics and enthusiasts about the issues raised-possi-
bly because of the different value systems involved. If one believes (with
Weizenbaum) that there are certain areas where computers should not be
used, this is an ethical, or perhaps a religious, judgment; and if one believes
(with Dreyfus) in a phenomenological approach to understanding, where
the feelings of the experiencing human body are central, one is committed
to an epistemological tradition foreign to virtually everyone in the world
of computer science and artificial intelligence.
(I ought to point out that Dreyfus's book went into a second edition
in 1979, and that he feels that certain trends in artificial intelligence-for
example, the adoption of organizing schemas or frames-go some distance
toward incorporating the human approach to experience. As for the com-
163
II I THE CoGNITIVE SciENCES : HisToRICAL PERSPECTIVE
puter science community, at least a minority now feels that Dreyfus has
raised issues that deserve to be addressed seriously. Nonetheless, Dreyfus's
next book is tentatively entitled Putting Computers in Their Place. )
Perhaps most disturbing to A.I. workers was the negative review given
to their field by a supposedly disinterested English observer, Sir James
Lighthill, who had been requested by his government's Science Research
Council to evaluate the state of the art in British artificial intelligence.
Lighthill found relatively little to admire and wrote disparagingly about
the distance between initial expectations and actual achievements in the
first twenty years:
When able and respected scientists write in letters to the present author that
AI, the major goal of computing science, represents "another step in the general
process of evolution"; that possibilities in the 1980s include an all-purpose intelli-
gence on a human-scale knowledge base; that awe-inspiring possibilities suggest
themselves based on machine intelligence exceeding human intelligence by the
year 2000 [one has the right to be skeptical]. (1972, p. 17}
Suffice it to say that programs already exist that can do things-or, at the very
least, appear to be beginning to do things-which ill-informed critics have asserted
164
Artificial Intelligence: The Expert Tool
a priori to be impossible. Examples include: perceiving in a holistic as opposed to
an atomistic way; using language creatively; translating sensibly from one language
to another by way of a language-neutral semantic representation; planning acts in
a broad and sketchy fashion, the details being decided only in execution; distin-
guishing between different species of emotional reaction according to the psycho-
logical context of the subject. (Boden 1981, p. 33)
Even as the sights have been lowered from the u absurd overoptimism" of
the 1950s, the actual achievements by certain workers are notable and have
convinced most observers outside of the field itself that the experiments
are at least interesting and should be taken seriously.
While the critics of A.I. were not easily silenced, there were, by general
consensus, a new burst of energy and several significant achievements in
the field in the early 1970s. This second wave is epitomized by Winograd's
SHRDLU, the shift from generalist to expert knowledge systems, and the
fusion of features of declarative and procedural approaches. At this time,
there was another vital, though rather controversial trend: the increasing
use of a top-down approach to the understanding of language and other
cognitive domains.
As exemplified in the work of Roger Schank and his colleagues at Yale
University (Schank 1972; Schank and Abelson 1977), an "understanding"
mechanism has several expectations of what a text is like in general; it also
incorporates a core set of knowledge structures about details of the subject
matter under discussion. These structures are built in as part of a prior
knowledge base; they can then be brought to bear upon a particular text
in an effort to comprehend how that text resembles, but also how it differs
from, other instances of its genre. In the best-known formulation, Schank
has introduced the notion of a script-the canonical set of events one can
expect in an often encountered setting, such as a meal at a restaurant or
a visit to a doctor's office. The script then allows one to make sense of
different meals, ranging from a snack at McDonald's to a banquet at
Maxim's; or of a series of visits to different medical specialists. Such a
structured framework allows the "understander" to deal expeditiously
with a variety of otherwise difficult-to-assimilate texts (much as Frederic
Bartlett's story schemas allowed his subjects to make sense of an otherwise
mysterious ghost story).
Another influential top-down approach is Marvin Minsky's (1975)
165
II I THE CoGNITivE SciENcEs : HISTORICAL PERSPECTIVE
Pluralisms
Minsky has also put forth an intriguing conception about how the
mind works, leading to novel proposals about how. computer programs
should be crafted. Rather than believing in a simple general processor or
central processor, through which all information must be passed, or in an
organized or unified mind that oversees all activity, Minsky now views
mental activity as consisting of many agents, each of which is a specialist
of some sort. According to this "society of minds" view, the mind consists
of several dozen processing centers, or "agents," which can handle differ-
ent types of knowledge simultaneously. Each of the local agents has a
function, which is called upon in certain circumstances, and each has access
to other agents. The agents communicate by emitting excitation and inhi-
bition rather than by transmitting symbolic expressions. They can also
censor information, much like a Freudian superego. Under this scheme,
some parts of the mind know certain things, while other parts know things
about the former. In fact, knowledge about which agents can know or
accomplish which things becomes a crucial component of this new way of
thinking about the mind (Minsky 1979, 1982).
Minsky's idea of a frame and his "society of minds" metaphor are not
in themselves theories which can be subjected to clear-cut scientific tests,
but are better thought of as organizing frameworks (frames, if you will)
which lead to the devising of programs that perform more effectively and
model human behavioral activity more faithfully. In that sense, his ideas
can be seen as reactions: reactions against those approaches that fail to
166
Artiftdal Intelligence: The Expert Tool
build in prior knowledge or expectations, and as well against those ap-
proaches that feature detailed knowledge of a specific area but have abso-
lutely no generality or connection to any other domain. The ultimate
impact of Minsky's new ideas on artificial intelligence is not yet known.
With these widespread shifts in how the knowledge base is conceptu-
alized, there has also been increasing dissatisfaction with the kind of
computer that usually serves the A.l. researcher: the serial digital "von
Neumann" computer. Scholars like Minsky himself, and younger ones like
Geoffrey Hinton and James Anderson (1981), raise an intriguing possibil-
ity: since the brain is itself a parallel rather than a serial mechanism, with
millions of neural events occurring simultaneously, the simulation of
human activities ought to be carried out by computers that also operate in
parallel fashion. The brain is an apparatus that learns, that executes many
special-purpose activities, and that has information dispersed throughout
large reverberating circuits. Hinton, Anderson, and their colleagues call for
a computer that more closely parallels the operation of the human brain,
and for programs that feature many cooperating individual agents. Most
of these efforts thus far involve the simulation of visual processing-an
area well enough understood in neurophysiological terms to permit a plau-
sible anatomical simulation. A growing number of experts speculate that
the next innovative wave of artificial intelligence will utilize "paraHel
kinds" of architecture for information processing (Feigenbaum and
McCorduck 1983). I shall take a look at some of these ideas about "non-
von Neumann" style computing in the discussion of visual perception in
chapter 10.
Understanding of Language
In addition to this influx of new ways of thinking, there have also
been impressive achievements in some of the specific subject areas of
artificial intelligence-achievements that reflect new concepts and can be
said to atone for the contrived performances of the earlier generation of
programs. In the area of language, for example, Roger Schank and his
colleagues at Yale University, as well as his students who have now moved
to other research centers, have produced programs that give precis of
stories, skim newspaper articles, and answer questions and draw inferences
about plot, character, and motivation in stories (Schank and Abelson 1977;
Wilensky 1983).
Schank's claims about his programs, as well as his more general theo-
ries of language, have generated much controversy. He traces the failure
of early language programs to their narrow focus on grammar and calls for
programs that truly "understand the language," or at least parts of it. But
167
I I I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
168
Artificial Intelligence: The Expert Tool
in other aspects of natural-language competence. Some approaches, such
as those of William Woods (1975) and Ross Quillian (1968), follow the
road of semantics; while others, such as those of Mitchell Marcus (1979),
prefer the trail of syntax. Raj Reddy's HEARSAY (Reddy et al. 1973)
program addresses the problem of understanding speech by drawing on
several kinds of knowledge-semantic, pragmatic, syntactic, lexical, pho-
nemic, and phonetic-and analyzing the signal at a variety of levels from
the single-sound segment to the complete phrase. Each knowledge source
takes a look at the hypothesis generated by all the others (and "displayed"
on a central blackboard) and then makes a "best" guess about what has
been said. While this system certainly lacks elegance, or the tightness of
a theoretically motivated approach, it exploits that quality of piecing to-
gether bits of evidence that may be involved in understanding speech
under less than optimal circumstances.
Perception
There has been analogous, and perhaps more unambiguous, progress
in the area of visual perception. Around 1960 a brief flurry of excitement
surrounded PERCEPTRON, a mechanism designed by Cornell's Frank
Rosenblatt for the purpose of recognizing letters (and other patterns)
placed in front of its "eyes." PERCEPTRON consisted of a grid of four-
hundred photocells, corresponding to neurons in the retina; these cells
were connected to associator elements, whose function was to collect elec-
trical impulses produced by the photocells; connections were made ran-
domly, because it seemed the best way to mimic the brain (Bernstein 1981;
Dreyfus 1972). (When random wiring did not work, PERCEPTRON was
rewired in a more deliberate way, which improved its performance in
pattern recognition.) Another set of components entailed response units:
an associator would produce a signal only if the stimulus was above a
certain threshold.
Marvin Minsky and Seymour Papert at M.I.T. believed that the rna-
chine was built upon erroneous concepts, and that it was more profitable
to find the principles that will make a machine learn than to build one in
the hope that it just might work. As they saw it, it was necessary to build
some prior structure in the machine and to provide the system with in-
formative feedback about successes or failures. Minsky and Papert eventu-
ally published a book in which they demonstrated conclusively the limita-
tions of PERCEPTRON theory (1968). For a while their critique put a
damper on the field, because A.I. workers drew the conclusion that nothing
further could be accomplished in the area of form recognition using sys-
tems based on the model of neural networks.
169
I I I THE CoGNITivE SciENCES : HisTORICAL PERSPECTIVE
170
ArHficial Intelligence: The Expert Tool
certain striking results won him widespread admiration before his un-
timely death at the age of thirty-five in 1980. I shall take a closer look at
Marr's accomplishments (Marr 1982) when I focus in chapter 10 on the
issue of how one recognizes an object.
Of course, many other researchers and programmers could be men-
tioned: both those involved in "mainline" work in language, visual percep-
tion, and problem solving; and those tilling peripheral soils, such as Chris-
topher Longuet-Higgins's (1979) work in the recognition of musical keys
or Richard Young's (1974) work simulating children's cognitive processes
on Piaget-type tasks. There are also under way promising new efforts,
whose overall impact cannot yet be judged. For example, Douglas Lenat
(1976; 1984) has developed an approach to heuristic reasoning as a supple-
mentary means for solving problems and even for finding new heuristics.
In any event, the general point should be clear: there has been a second
burst of energy in the field of artificial intelligence since the crisis of the
early 1970s. If the accomplishments are not yet up to human snuff, they
can no longer be readily dismissed as resting on superficial procedures or
depending on cheap tricks, but are clearly addressing central issues in
human intelligent behavior.
171
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
Searles Conundrum
To combat the strong view of A.l., Searle introduces his provocative
Chinese room example. Suppose, he says, that an individual (in this case,
Searle) is locked in a room and given a large batch of Chinese writing.
Suppose that he knows no Chinese writing and may not even be able to
discriminate Chinese from other kinds of squiggles. Suppose next that he
is given a second set of Chinese characters together with a set of rules for
collating the second batch with the first. The rules are in English and are
therefore understood by Searle. The rules teach him to correlate one set of
formal symbols with another set. This process is repeated with other
materials so that the speaker becomes accomplished at correlating the
characters with one another and, hence, can always provide the "right" set
of characters when given an initiating set. Then, to complicate the situation
further, suppose that the speaker is also given questions and answers in
English, which he is also able to handle. The speaker is then faced with
the following situation:
Suppose also that after a while I get so good at following the instructions for
manipulating the Chinese symbols and the programmers get so good at writing the
programs that from the external point of view-that is, from the point of view of
somebody outside the room in which I am locked-my answers to the questions
are absolutely indistinguishable from those of native Chinese speakers. Nobody
just looking at my answers can tell that I don't speak a word of Chinese. Let us
also suppose that my answers to the English questions are, as they no doubt would
be, indistinguishable from those of other native English speakers, for the simple
reason that I am a native English speaker. From the external point of view-from
the point of view of someone reading my "answers"-the answers to the Chinese
questions and the English questions are equally good. But in the Chinese case,
unlike the English case, I produce the answers by manipulating uninterpreted
formal symbols. As far as the Chinese is concerned, I simply behave like a com-
puter: I perform computational operations on formally specific elements. For the
purposes of the Chinese, I am simply an instantiation of the computer program.
(Quoted in Hofstadter and Dennett 1981, p. 33)
Searle's argument is more complex and subtle than this single quota-
tion reveals, but its general tenor can be grasped. He believes that the
Turing test is no test whatsoever of whether a computer program has a
mind or is anything like a human being, because humanlike performance
can be faked by any individual or machine that has been supplied with a
set of formal rules to follow under specific circumstances. So long as one
172
Artificial Intelligence: The Expert Tool
is simply following rules-so long as one is engaged in syntactic operations
-one cannot be said to truly understand. And the computer is a machine
par excellence for formal operations, unencumbered by any semantic knowl-
edge, by any knowledge of the real world, or by any intention to achieve
certain effects by its specific response. Therefore, it is a fundamentally
different kind of entity from a human being who understands the semantic
content of an utterance, and has his own purposes for communicating.
Conclusion: the program of strong A.I. is bankrupt.
Counterattacks
Knowing that his argument will not go unchallenged, Searle himself
anticipates the major counterarguments. He reviews the Berkeley or "sys-
tems reply": the claim that while the individual in the room may not
understand the story, he is part of a whole system-ledger, paper, data
banks, and so on; and the system itself understands. Searle's response is
that the individual could memorize or internalize all the materials in the
system (for example, memorize the rules in the ledger and the data lists of
Chinese symbols and do all calculations in his head) and still would not
understand.
A second response is the robot response of the Yale Contingent.
According to this line of argument, a computer could be put inside a robot,
and the robot could walk around and do all the sorts of things that real
people do; being thus in touch with the real world, such a robot would
really understand. But Searle says that, while this reply tacitly agrees that
understanding is more than formal symbol manipulation, it still fails to
make its case-because Searle could, unbeknownst to himself, actually
become the robot. For example, the Chinese symbols might actually come
to him from a television camera attached to the robot: thus, the Searle-
robot would appear to be perceiving. Analogously, the symbols that he is
issuing forth could make motors move the robots, arms, and legs: thus, the
Searle-robot would act. In either case, however, Searle would actually be
making all of the formal manipulations of symbols by himself and still
would know nothing of what is going on. In one sense, he would be the
homunculus of the robot but would remain in total ignorance of the
Chinese meanings.
A third line of argument, traced this time to Berkeley and M.I.T., is
the brain-simulator reply. This response suggests another scenario: a pro-
gram now simulates the actual sequence of neuron firings at the synapses
of the brain in a native Chinese speaker when he is engaged in, for exam-
ple, story understanding or question answering. The machine now taking
in stories simulates the formal structure of actual Chinese brains in proc-
173
II I THE CoGNITivE SciENCES : HisTORICAL PERSPECTIVE
essing the stories, and gives out Chinese answers as output. Searle says that
even creeping up this close to the operation of the brain still does not
suffice to produce understanding. Instead, he says, imagine that rather than
a monolingual man being in a room shuffling symbols, the man is equipped
with an elaborate set of water pipes with valves connected to them. When
the man receives the symbols, he looks them up in the program and then
determines which valves to turn on and off. Each water connection corre-
sponds to a synapse in the Chinese brain, and the whole system is so rigged
that eventually Chinese answers pop out. Now, points out Searle, the man
does not understand Chinese, nor do the water pipes. The problem with
the brain simulator is that it is simulating the wrong things about the brain.
Searle indicates that
174
Artificial Intelligmce: The Expert Tool
standing were it to be found in a human being. There are orders of magni-
tude Searle ignored (for example, the effort it would take to anticipate and
translate all possible messages) and only this sleight-of-hand allows him
to make his case. Searle has also been accused of confusing levels: it is the
person who understands, and not the brain; and similarly, it is the com-
puter as a whole that understands, and not the program.
To my mind, when Searle remains on his own ground-in putting
forth the example of the man equipped with a set of formal rules for
responding in Chinese-he is convincing. While I would not join Searle in
ruling out the possibility of any machine's understanding a natural lan-
guage, I certainly feel that the word understand has been unduly stretched
in the case of the Chinese room problem in its various manifestations.
It is Searle's positive explanation of why the rest of us understand,
when he speaks about the intentionality of the brain, that can be legiti-
mately attacked. The whole notion of the brain as a causal system that
displays intentionality is obscure and difficult to understand, let alone to
lay out coolly like a computer program. Zenon Pylyshyn, a computer
scientist from the University of Western Ontario, scores telling points:
"What Searle wants to claim is that only systems that are equivalent to
humans ... can have intentionality. His thesis thus hangs on the assump-
tion that intentionality is tied very closely to specific material properties
-indeed, that it is literally caused by them." Pylyshyn wonders whether
Searle is proposing that intentionality is a substance secreted by the brain
(1980a ), and then poses this puzzle:
[Suppose] if more and more of the cells in your brain were to be replaced by
integrated circuit chips, programmed in such a way as to keep the input-out
function of each unit identical to that of the unit being replaced, you would in all
likelihood just keep right on speaking exactly as you are doing now except that
you would eventually stop meaning anything by it. What we outside observers
might take to be words would become for you just certain noises that circuits
caused you to make. (P. 442)
175
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
Indeed, Searle seems to believe that, to do cognitive studies, one needs only
two levels of explanation-the level of intentionality (a plain English
discussion of the organism's wishes, beliefs, and so on)-and a neuro-
physiological explanation of what the brain does in realizing these inten-
tional states. He finds no need for a level of symbolic representation, which
dominates work throughout the cognitive sciences (1983b). To the extent
that these lines of argument sound like mere handwaving, Searle's positive
assertions about intentionality have little force. To the extent that he can
specify just what he means, and show that intentionality is not in principle
restricted to organic brains, a genuine scientific issue will have been joined.
Thereafter, researchers in artificial intelligence can try to simulate the very
behavior that Searle believes lies beyond their explanatory scope.
It is possible, however, that the issues between Searle and his critics
-and perhaps even between those in sympathy and those out of sympa-
thy with A.I.-transcend questions of a scientific nature. In an angry
critique of the Searle paper, computer scientist Douglas Hofstadter makes
the following conjecture:
176
Artificial Intelligence: The Expert Tool
is a matter of faith, in which reason will not even be considered relevant
by either side.
Whatever the merit of Searle's arguments, they do not question the legiti-
macy of most current work in artificial intelligence; in fact, as he indicates,
he has no qualms about accepting the utility of A.l. as an aid in conceptual-
izing and testing theories of human intelligence. But, as we have seen,
some critics challenge A.I. at an even more fundamental level. As one
example, Joseph Weizenbaum (1976), raises ethical questions about
whether A.l. should be allowed to impinge on territory that has hitherto
been restricted to human beings, such as issues of justice or love.
Another line of criticism suggests that efforts to simulate human intel-
ligence are perfectly valid, but that most of the A .I. community has hereto-
fore used superficial models which do not approach the core of human
thought processes. Initially, this line of criticism came from those in the
camp of transformational linguistics, like Elan Dresher, Norbert Hornstein
(Dresher and Hornstein 1976), and Kenneth Wexler (1978). From their
perspective, rather than trying to build upon deep principles of language
(or whatever area happens to be modeled), artificial intelligence is simply
a practical, trial-and-error pursuit and is unlikely ever to come up with
anything that has generality and explanatory power. For instance, as they
see it, Schank's program is too vague and unspecified, and there are no
algorithms for determining the "primitive verb" or the script that is rele-
vant to a given linguistic string. In contrast, Winograd's program is criti-
cized for being too specifically tied to a given "micro-world" and thus
being inapplicable to any new specimen of language. According to this
"purist" philosophy, scientists ought to be spending their time analyzing
systems in themselves, studying competent human beings, or fashioning
theoretical explanations of what intelligence or competence amounts to:
theories that can stand up to the standards of the "hard" sciences. Perhaps,
once the real operative principles are understood-as they are beginning to
be understood in the area of syntactic processing-it should be relatively
trivial to program the computer to carry out these operations. In the next
chapter, I shall examine this transformationalist vision of acceptable ex-
planatory models in cognitive science.
These critics are arguing that artificial intelligence has not yet attained
the standing of a genuine scientific endeavor. Instead, it is just a convenient
177
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
Current systems, even the best ones, often resemble a house of cards .... The
result is an extremely fragile structure which may reach impressive heights, but
collapses immediately if swayed in the slightest from the specific domain (often
even the specific examples) for which it was built. (Bobrow and Winograd 1977,
p. 4)
AI is very hard. What is the nature of knowledge? How do you abstract from
existing knowledge to more general knowledge? How do you modify knowledge
when you fail? Are there principles of problem solving that are independent of
domain? How do goals and plans relate to understanding? The computer is a way
of testing out ideas. But first, we need to understand what we're supposed to be
building models of. (Quoted in Waldrop 1984a, p. 805)
Workers like John McCarthy (1984) have been pondering what Daniel
Dennett (1983) terms the "smoking gun" problem of artificial intelligence:
how a program can decide which of the unexpected features in a situation
can be ignored. (For example, if it is to resemble human intelligence, a
frame of a boat trip should ignore the fact that the oars on a rowboat are
blue rather than green but dare not ignore the fact that one of the oarlocks
is broken.) John Seely Brown and Kurt van Lehn, researchers at Xerox
P ARC in Palo Alto, admit that most current A. I. work fails to meet tradi-
tional criteria of scientific theories, and decry the absence of "competitive
argumentation" whereby the power of one simulation can be rigorously
compared with the power of another. By such comparison, it is possible
to indicate why one accepts some computational principles, while rejecting
others. In lieu of such argumentation, these authors wryly suggest, artifi-
cial intelligence theories have stood on the toes rather than on the shoul-
ders of their predecessors (van Lehn, Seely Brown, and Greeno 1982).
The realization that one's chosen discipline may not have been operat-
ing in a scientifically impeccable manner is not in itself an instant panacea,
but it is an important first step-maybe the most important step in the
maturing of a discipline. A second valuable trend, in my view, has been
the growing number of workers in A.I. who seek systems reflecting the
178
Artificial Intelligence: The Expert Tool
deep principles at work in a particular domain of knowledge. Spurning
both the search for the most general properties of problem solving (which
proved elusive with the General Problem Solver) and the interest in those
expert systems that perform well simply by drawing with brute force on
massive amounts of stored knowledge, these researchers take seriously the
principles that appear to be at work in the only other intelligent system
we know-the human being. I shall examine some of their promising
efforts in the latter chapters of this book.
It is possible, of course, as a result of a careful study of different areas,
we will discover that the deepest principles do indeed operate across areas,
and that we will ultimately require a general knowledge base or problem
solver rather than a host of separate experts. As I suggested with respect
to the work of John Anderson (1983) in cognitive psychology, such a
dream may still materialize, but the likelihood seems small.
My own analysis suggests that, after a period of excessive claims and
sometimes superficial demonstrations, artificial intelligence has advanced
to a more measured view of itself and has, in the process, attained a set of
reasonably solid accomplishments. This maturing process has involved a
recognition that the practice of A.I. entails deep philosophical issues which
cannot be ignored or minimized. Involved as well has been a recognition
that there are limits to what can be explained by current A.I. methods and
that even whole areas of study may lie outside of artificial intelligence, at
least now and perhaps permanently. It is important to have genuine demon-
strations and not just verbal descriptions of possible programs: this insis-
tence has been among the greatest contributions of Newell and Simon. But
it is equally important that these demonstrations reflect robust principles
and are not just fragile constructions with limited application.
Given my own view that the future of cognitive science lies in the
study and elucidation of particular domains of knowledge, I believe that
the field will achieve scientific status when it freshly illuminates the do-
mains and knowledge systems with which it is concerned. As the psy-
chologist John Marshall declares:
I just don't care what members of the AI community think about the ontologi-
cal status of their creations. What I do care about is whether anyone can produce
principled, revealing accounts of, say, the perception of tonal music ... , the
properties of stereo vision ... , and the parsing of natural language sentences.
Everyone that I know who tinkers around with computers does so because he has
an attractive theory of some psychological capacity and wishes to explore certain
consequences of the theory algorithmically. (1980, p. 436)
179
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
180
Artificial Intelligence: The Expert Tool
central components of a new cognitive science, but they cannot in them-
selves constitute the field. Philosophy will remain a source of important
questions, a repository of invaluable reflections on those questions, and a
valuable critic of cognitive scientific practices. But an equally important
partner in this newly fashioned line of inquiry will be those disciplines that
take as their appointed task the analysis of particular domains of cognition.
Of the various disciplines that must assume this burden, linguistics is, by
all accounts, the discipline that has accomplished the most and is thus most
centrally involved in cognitive science. Moreover, and perhaps for this
reason, other disciplines look to linguistics as a possible model of how to
study a particular domain and to apply the powerful (and perhaps too
powerful) tools of the psychologist and the computer scientist. Because of
the intrinsic importance of language among the human cognitive faculties,
because of linguistics' role as a possible model for domain-specific studies,
and because of the tremendous strides made in linguistics over the past few
decades, it is appropriate to turn now to the scientific study of language.
181
7
Enigmatic Sentences
At first reading, one of the most famous sentences of twentieth-
century science makes little everyday sense. This is Noam Chomsky's
oxymoronic proposition "Colorless green ideas sleep furiously." After all,
ideas don't have colors, ideas don't sleep, something can't be simultane-
ously colorless and green-and for that matter, what does it mean for any
entity to sleep furiously?
And yet a certain sense can be squeezed out of this sentence. It is
certainly clearer than "ideas furiously green colorless sleep" or some other
random concatenation of these five substantives. In Chomsky's sentence,
one knows that a proposition is being asserted about ideas; one senses a
contrast being drawn between an apparent state of blandness, as against
a more active state of energy and color. The sentence is even susceptible
to a figurative reading. One thinks of the revolutionary impact of appar-
ently innocuous ideas-the kinds of idea Noam Chomsky himself put
forth in his 1957 monograph Syntactic Structures.
The fact that we have clear intuitions about apparent nonsense like
"Colorless green ideas sleep furiously" (or "Twas brillig and the slithy
toves did gyre and gimble") underlies Chomsky's central contributions to
linguistics and, for that matter, to cognitive science in a broader sense.
182
Linguistics: The Search for Autonomy
What Chomsky accomplished in his monograph, and in the numerous
writings that followed, was to call attention to certain properties of sent-
ences which all normal speakers and hearers intuitively know, but which
derive from a deeper understanding of the language whose properties may
be explicitly known only to the linguist. Chomsky called attention to and
offered a convincing explanation of differences between apparently similar
sentences: "John is easy to please" (where John is the recipient of the
pleasing) versus "John is eager to please" (where John does the pleasing).
"Ted persuaded John to learn" (where John is to do the learning) versus
"Ted promised John to learn" (where Ted does the learning). Chomsky
provided an account of sentences where meaning can be preserved despite
shifting of the principal terms ("The cat chased the mouse" versus "The
mouse was chased by the cat") as against those sentence pairs where one
cannot simply reverse clauses ("Many men read few books" versus "Few
books are read by many men"). He singled out and suggested mechanisms
underlying the human ability to detect and unravel the ambiguities con-
tained in such sentences as "Flying planes can be dangerous," "The shoot-
ing of the hunters disturbed me," and "I didn't shoot John because I like
him." And he marked for study sentences that seem acceptable: "John
seemed to each of the men to like the other" as against those superficially
similar sentences that violate some putative rule of the language: "John
seems to the men to like each other." We can say "We like each other"
meaning each of us likes the other, but we can't say "We expect John to
like each other"-indeed we're not even sure what this could mean
(Marshall 1981).
That human beings have intuitions for sentences like these was not
an insight unique to Chomsky. After all, Lewis Carroll exploited similar
knowledge in fabberwoclcy. The influential linguist of the early part of this
century, Edward Sapir (1921), had called attention to many of the very
relationships among sente~ces that were later to occupy Chomsky. But
Chomsky went beyond his predecessors, including insightful linguists like
Sapir, in his goal of setting out the rules that allow individuals to make,
or to generate, all of the correct sentences listed above; to know that they
are correct and what they mean; and to be able to pick out those sentences
that violate these rules and are hence ungrammatical, though not necessar-
ily devoid of meaning. To do this, the speaker must possess at some
level a detailed set of rules or procedures indicating when different parts of
speech can occur in given places in an utterance: that is, the rules must
capture the intuitions of native speakers about the relations obtaining
within and among sentences. Chomsky's statement of this explicit goal for
the study of language-or, more precisely, for the study of syntax-and
his success in developing methods directed toward the achievement of this
183
I I I THE CoGNITIVE SciENCES : HisToRICAL PERSPECTIVE
goal instantly made his work stand out from that of other students of
language. And his more general conviction that the several domains of
mind (such as language) operate in terms of rules or principles that can be
ferreted out and stated formally constitutes his main challenge to contem-
porary cognitive science.
In Syntactic Structures, Chomsky first announced a set of goals and then
described the kind of grammar that appeared necessary if the proper
regularities in English-or, indeed, in any language-were to be discerned.
Rather than simply looking at the data of language, and trying to discern
regularities from empirically observed utterances, as his predecessors had
typically done, Chomsky insisted that the principles would never emerge
from a study of the utterances themselves. Instead, it was necessary to
work deductively. One must figure out what kind of a system language is,
much in the way that one figures out what a particular branch of mathe-
matics is like; and one must state one's conclusions in terms of a formal
system. Such an analysis should lead to positing of the rules that can
account for the production of any conceivable grammatical sentence (and
there is, of course, an infinite number of such sentences), but at the same
time the rules should not "generate" any incorrect or ungrammatical sen-
tences. Once the system has been set up, one should then examine particu-
lar utterances to determine whether they can, in fact, be appropriately
generated through adherence to the rules of the linguistic system.
In embarking on this program, Chomsky made two important, simpli-
fying assumptions. One assumption was that the syntax of language could
be examined independently of other aspects of language. If syntax were
inextricably bound up with other aspects of language-for example, with
meaning or with communicative utility-then it might not be possible to
figure out its governing laws. The second but closely related tacit assump-
tion was that the discipline of linguistics could proceed independently of
other areas of the cognitive sciences. Once again, if the study of language
were integrally tied to the study of other areas of human cognition-to
other mental organs, as Chomsky would later phrase it-then progress
might be impossible or agonizingly slow.
For the most part, these working assumptions about linguistic auton-
omy-the autonomy both of syntax from other aspects of language, and
of linguistics from other aspects of cognitive science-worked out propi-
tiously and made linguistics a rapidly developing area of science. But
whether these assumptions can ultimately be sustained-whether the as-
sumptions of autonomy are truly justified-constitutes a problem that has
yet to be resolved. I shall return to these issues in the concluding section,
after reviewing Chomsky's principal discoveries, the history of the field
prior to his own entry, and certain current issues in the science of language.
184
Linguistics: The Search for Autonomy
Before entering into the heart of Chomsky's contributions, it may be
well to say a word about the focus of this chapter. While the work of other
scholars has been central in other chapters, in no other chapter of this book
has so much attention been focused on a single individual. In part, this is
an expositional device-a way of presenting the principal (and often com-
plicated) ideas of modern linguistics in as accessible a fashion as possible.
But, also, in no other contemporary cognitive science is the work of a single
individual so key and so irreplaceable. In a nontrivial sense, the history of
modern linguistics is the history of Chomsky's ideas and of the diverse
reactions to them on the part of the community.
s
Chomsky Approach
To achieve his ambitious goal, Chomsky first had to show that current
methods for analyzing syntax and accounting for acceptable sentences
would not work. By pushing a precise but inadequate formulation to its
unacceptable conclusion, one could show why that formulation was inade-
quate-and, in the process, obtain a deeper understanding of the linguistic
data it had vainly sought to explain. Chomsky proceeded by proving that
the theoretically most plausible method for generating sentences could not
in principle work. To begin with, he considered a finite-state grammar: a
machine with a finite number of internal states which generates sentences
in the following manner. Starting in a unique initial state, the machine
passes into a second state by producing the first word of a sentence; then,
while emitting a word at each transition, the machine proceeds from state
to state until it has reached the final state, at which time a complete
sentence has been generated. Chomsky showed that such finite-state
grammars are inherently incapable of representing the recursive properties
of English constructions: that is, by its nature, a finite-state grammar
cannot generate sentences in which one clause is embedded in or depen-
dent upon another, while simultaneously excluding strings that contradict
these dependencies.
As an example, consider the sentence "The man who said he would
help us is arriving today." Finite-state grammars cannot capture the struc-
tural link between man and is arriving which spans the intervening relative
clause. Moreover, as its name suggests, a finite-state grammar cannot han-
dle linguistic structures that can recur indefinitely, such as the embedding
of a clause within another clause ("the boy that the girl that the dog
... ," and so on). Even though such sentences soon become unwieldy for
the perceiver, they are strictly speaking grammatical: any grammar must
be able to account for (or generate) them. At the most general level, English
(and, for that matter, other languages) does not work by slotting a word
185
II I THE CoGNITivE SciENCES : HisTORICAL PERSPECTIVE
and then indicating what words can follow "to the right" of it; it works
at a higher level of abstraction, where certain elements, under certain
circumstances, can be placed wholly within other elements. Thus,
Chomsky was making (with respect to language) precisely the point that
Karl Lashley had urged about serial behavior in all of its manifestations.
Chomsky also demonstrated the unwieldiness-though not the utter
impossibility--of a second kind of grammar, a phrase-structure grammar (or an
immediate constituent analysis). Here the linguist works with an initial set
of strings, together with a finite set of phrase-structure or "rewrite rules,"
where one phrase can be rewritten in another permissible form. Chomsky
showed that such a phrase-structure system can only generate with great
unwieldiness certain sentences; moreover, it cannot capture or account for
many of the regularities any English speaker appreciates. Thus a phrase-
structure grammar cannot explain the different structures of "What are
you looking for" and "What are you running for" (Lees 1957, p. 386), or
the ambiguity of "The teacher's marks are very low." Nor is there any way
of showing that "John hit Bill" and "Bill is hit by John" bear a close
relationship to one another. Nor, according to Chomsky, is it possible,
using a phrase structure grammar, to generate sentences that involve the
combination of two parallel clauses: to be specific, such grammars offer no
mechanism for combining sentences like "The scene of the movie was in
Chicago" and "The scene of the play was in Chicago" into "The scene of
the movie and of the play was in Chicago." In broader terms, the phrase-
structure grammar can be written only at the cost of several restatements
of extensive parts of the grammar. As linguist Robert Lees points out, "if
these uneconomical repetitions are permitted, then the grammar fails to
state the near identity in structure between those parts which must be
repeated" (1957, p. 387). More generally, while English might conceivably
be described in terms of phrase-structure grammars, such a description
would be so unwieldy, so hopelessly complex that these descriptions
would be of scant interest.
Thus, the most plausible models for the generation of grammatical
sentences were both shown to be inadequate: finite-state grammars, be-
cause countless instances of meaningful language cannot be generated on
a word-by-word basis; phrase-structure grammars, because mere attention
to the ways that phrases are constructed fails to capture important
regularities in the language. For these reasons, Chomsky found it necessary
to introduce a new level of linguistic structures which at once swept away
these difficulties and made it possible to account for the full range of
sentences in the language.
Inspired in part by his teacher Zellig Harris (1952), Chomsky discov-
ered the level of transformational analysis. In a transformational grammar,
186
Linguistics: The Search for Autonomy
Chomsky argued, one posits a set of rules whereby sentences can be related
to one another, and where one sentence (or more precisely, the abstract
representation of a sentence) can be converted or transformed into another.
A generative grammar, in Chomsky's sense, is a rule system formalized
with mathematical precision: without drawing upon any information that
is not represented explicitly within it, the system generates the grammati-
cal sentences of the language that it describes or characterizes, and assigns
to each sentence a structural description or grammatical analysis.
As described in Chomsky's 1957 monograph, the method of transfor-
mational grammar works roughly as follows. (It must be immediately
added-indeed, stressed-that the system has changed several times in the
intervening years.) Beginning with phrase-structure rules, one generates
only the cores of sentences, or kernel sentences, which are short active declar-
ative assertions. These are generated by following a set of instructions for
constructing strings: for example, (1) Sentence = Noun Phrase + Verb
Phrase; (2) Noun Phrase = T + N; (3) T = the; (4) N = man, ball; (5)
Verb Phrase = Verb + Noun Phrase; (6) Verb = hit, take; and so on.
Starting with the single symbol S, one can generate, by a completely
specifiable set of rules, a kernel sentence like "The man hit the ball."
Thereafter, all the other grammatical sentences of the language can be
generated by means of transforming these kernel sentences.
Transformations are an algorithmic set of procedures that occur in a
prescribed order and allow one to convert one linguistic string to another.
Thus, a transformation allows one to convert an active sentence into a
passive sentence, a positive statement into a negative statement or a ques-
tion. On such an analysis, "What are you looking for" can be described
as the "what-question" transformation of "You are looking for it"; while
"What are you running for" is a "why-question" transformation of "You
are running." Transformations can lay bare the links between sentences
like "The boy kissed the girl" and "The girl was kissed by the boy"; the
deep differences between superficially similar syntactic arrangements like
"The girl is eager to please" and "The girl is easy to please"; and the fact
that a phrase like "the shooting of the hunters" is ambiguous because
it can be accounted for by two different transformational histories-
one, in which the hunters are shot; the other, in which the hunters do the
shooting.
All these transformations are structure-dependent: that is, they do not
operate on single words or on strings of words of arbitrary length. Rather,
the transformations are imposed on strings (abstract representations of
sentences) after these strings have been analyzed into appropriate syntactic
categories and constituents (or phrase structures), which determine when
and where the transformations can be applied. Because Chomsky placed
187
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
188
Linguistics: The Search for Autonomy
Chomsky also characterized the task of the linguist in a more explicit
way than had his predecessors. He devoted considerable attention to the
theoretical question of how one goes about choosing one linguistic model
over another. He laid out the formal criteria for an adequate theory of
linguistics suggesting (and demonstrating) how these criteria might be
achieved. He expounded as well an ordered set of standards of adequacy:
observational adequacy, where the grammar gives a correct account of observed
linguistic data; descriptive adequacy, where the account also captures the na-
tive speaker's intrinsic competence, his own internalized knowledge; and
explanatory adequacy, where the analyst uncovers principles underlying opti-
mal grammar for the language. An account that exhibits explanatory ade-
quacy can explain why individuals have constructed the grammars that
they have, and what kinds of abilities individuals have needed to achieve
linguistic competence. Such theoretical sophistication (and presumption!)
had also been hitherto absent from the writings of linguists interested in
grammar.
189
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
Lees said that Syntactic Structures was one of the first serious attempts
to construct a comprehensive theory of language; but, as Frederic New-
meyer declares in his useful history of this era, "Actually the tone of the
review as a whole made it clear that Lees regarded it as the ONLY serious
attempt and a completely successful one at that" (1980, p. 19).
As they became aware of Chomsky's work, other linguists also dis-
cerned its power and originality. In another review, C. F. Voegelin, a
leading linguist, declared that even if Syntactic Structures managed to achieve
only part of its goals, "it will have accomplished a Copernican revolution"
(quoted in Newmeyer 1980, p. 19). The distinguished British linguist C. E.
Balzell remarked that "linguistics will never be the same"; and Bernard
Bloch, one of the leading linguists of an earlier generation, confided to
colleagues "Chomsky really seems to be on the right track. If I were
younger, I'd be on his bandwagon too" (both quoted in Newmeyer 1980,
p. 47).
But in heralding a revolution, Chomsky was launching a movement
at someone's expense-namely, the structural linguists of an earlier gener-
ation. And, as the full dimensions of his program became clear, there was
considerable and often violent opposition. For it turned out that Chomsky
was not merely offering a critique of existing attempts to describe a gram-
mar; instead, he espoused radically different views of how linguistics
should work, and, indeed, of what social science (we would today substi-
tute "cognitive science") was about. Some of these views were muted, or
had not yet been well formed, at the time of the publication of Syntactic
Structures (Chomsky was only twenty-nine at the time of the book's publi-
cation; twenty-seven at the conclusion of his doctorate); but with the
publication of succeeding articles in the late 1950s and early 1960s, and the
issuing of a major theoretical statement, Aspects of a Theory of Syntax, in 1965,
the full dimensions of his program had been laid out.
To begin with, Chomsky made it clear that he was interested in
language in a particular sense. Language is not a general means of commu-
190
Linguistics: The Search for Autonomy
nication or a blanket word encompassing all symbol systems: rather, at the
core of language is the property of syntax, the capacity unique to the
human species to combine and recombine verbal symbols in certain spe-
cifiable orders, in order to create a potentially infinite number of grammati-
cally acceptable sentences. Moreover, syntax was seen as the primary,
basic, or (in a technical sense) deep level of the language, with both seman-
tics (meaning) and phonology (sound structure) being constructed upon a
syntactic core. Thus, in a transformational account, the semantic and
phonological components have access to the output of the core syntactic
component, but not vice versa (Smith and Wilson 1979, p. 97): indeed, in
such an account, the verbal symbols are manipulated without any refer-
ence to their meanings or their sounds. Recently Chomsky (1981) has
nominated the shift from "language" to "grammar" in a restricted sense
as one of the two most important shifts in linguistic theory over the past
few decades (the other being the search for a more restricted notion of
grammars, so that possibilities for explanatory adequacy are enhanced).
Chomsky was also interested in language in a more abstract sense than
his predecessors had been. While the structuralists of the preceding gener-
ation examined what individuals actually said, and typically focused on
uttered words, Chomsky considered language to be an abstraction-a ca-
pacity that can merely be glimpsed in impure form in an individual's actual
output. Consistent with his formalistic leanings, Chomsky felt that lin-
guists should focus on this idealized, virtually Platonic form and disregard
the slight errors, pauses, lapses of memory, and the like which are as-
sociated with actual speaking in the real world. Both in his emphasis on
pure syntax and in his desire to study language as an ideal form, Chomsky
was dramatically at odds with the behaviorism of the time.
191
I I I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
free to talk about whatever one wants, in the way that one wants to.
Stressing the infinite expressive potential of language, Chomsky argued
that it is senseless to think of linguistic output as restricted in any mean-
ingful way by the stimuli that happen to be present.
With relish, Chomsky showed that Skinner's attempts to explain
language along these stimulus-response lines were fundamentally flawed.
They failed whether they were applied literally or metaphorically. As
Chomsky put it with characteristic asperity:
a critical account of his book must show that with a literal reading (where the
terms of the descriptive system have something like the technical meanings given
in Skinner's definitions) the book covers almost no aspects of linguistic behavior,
and that with a metaphoric reading, it is no more scientific than the traditional
approaches to this subject matter, and rarely as clear and careful. (1964a, p. 552)
But Chomsky made it clear that Skinner was also on the wrong track
epistemologically. Like other empiricists of the day, Skinner had urged
investigators to remain close to the data and to spurn abstract theory.
Chomsky, on the contrary, felt that the data would never speak for them-
selves, that it was necessary to take a theoretical stand and to explore the
consequences of that theory. Moreover, he revealed his own suspicion that
the kinds of theories needed to explain language, and other aspects of
human thought and behavior, would have to be abstract and, in fact,
frankly mentalistic. Leaning explicitly on the work of Descartes some three
hundred years earlier, and borrowing leaves as well from the writings of
Plato and Immanuel Kant, Chomsky argued that our interpretation of the
world is based on representational systems that derive from the structure
of the mind itself and do not mirror in any direct way the form of the
external world. And, indeed, once linguists accepted the impossibility of
a physical definition of grammaticality-for example, one expressed in
terms of actual utterances-they would realize the necessity for abstract
mentalistic linguistic theory.
In this book review, Chomsky also revealed his fundamental impa-
tience with most psychological approaches. In his view, psychological
experiments were often unnecessary to prove a point: many psychologists
preferred demonstrating the obvious to thinking systematically and in a
theoretical vein (Rieber 1983, p. 48). One psychologist Chomsky singled
out for praise was Karl Lashley. In his study of serially ordered behavior,
Lashley had concluded that an utterance is not simply produced by string-
ing together a sequence of responses under the control of outside stimula-
tion and interverbal association, and that the syntactic organization of an
utterance is not something directly represented in the physical structure of
the utterance. Rather, syntactic structure turns out to be, in Lashley's
192
Linguistics: The Search for Autonomy
words, "a generalized pattern imposed on the specific acts as they occur"
(quoted in Jeffress 1951, p. 122). Building on Lashley, Chomsky concluded
that there must be "a multiplicity of integrative processes which can only
be inferred from the final results of their activity" (1964a, p. 575).
A new psychological perspective appeared latent in Chomsky's for-
mulation. At the very end of the review, he opined, "At the moment the
question cannot be seriously posed, but in principle it may be possible to
study the problem of determining what the built-in structure of an infor-
mation-processing (hypothesis-forming) system must be to enable it to
arrive at the grammar of a language from the available data in the available
time" (1964a, p. 578). In the years ahead, Chomsky was to mine this
intuition, suggesting that the individual is born with a strong penchant to
learn language, and that the possible forms of the language which one can
learn are sharply limited by one's species membership with its peculiar
genetic inheritance. Chomsky was impressed by the extreme abstractness
of the task faced by every child who must learn language, the rapidity with
which the language is learned, despite the lack of explicit tutelage. Skin-
ner's notion of learning language through imitation and reinforcement was
implausible-if not impossible in the same sense as the finite-state gram-
mar: an entirely new psychological perspective was needed to account for
the rapidity and relatively error-free way in which children acquire lan-
guage, despite the "poverty of stimulus," the relatively small corpus of
often incomplete or error-tinged utterances they encounter in daily life.
193
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
the like. (In this move, he was reflecting, if not anticipating, similar shifts
occurring in psychology and artificial intelligence.) He came to think of the
mind as a series of relatively independent mental organs or modules, each
with its own rules of operation and development, along with prescribed
means of interacting with other "organs." Here was a powerful statement
on behalf of the autonomy both of language as an organ-and, not coinci-
dentally, of linguistics as a discipline. Superimposed on this modularity was
a commitment to mentalism-the existence of abstract structures in the
mind which make knowledge possible. And there was as well a swing to
nativism, the belief that much of our knowledge is universal and inborn, and
that, without the need for tutelage, one has access to this knowledge
simply by virtue of being human. These views were as much a departure
in the human sciences of the era as were Chomsky's initial claims in the
area of linguistics; and they were to generate even more controversy than
his initial views expressed in Syntactic Structures.
The full force of Chomsky's challenge to traditional linguistics be-
came clear in 1958 at the Third Texas Conference on Problems of Linguistic
Analysis of English. Here, in initially respectful but unfailingly firm terms,
he told his linguistic brethren that their traditional approaches to under-
standing language were doomed to fail. He asserted that a complete inven-
tory of elements in language could never give rise to a characterization of
all possible sentences, and that an inductive discovery procedure could
never work. He then went on to argue:
I think that the failure to offer a precise account of the notion "grammar" is
not just a superficial defect in linguistic theory that can be remedied by adding one
more definition. It seems to me that until this notion is clarified, no part of linguistic
theory can achieve anything like a satisfactory development. . . . I have been
discussing a grammar of a particular language here as analogous to a particular
scientific theory, dealing with its subject matter (the set of sentences of this lan-
guage) much as embryology or physics deals with its subject matter. (1964b, p. 213)
Chomsky took issue with the view that the methodological burden of
linguistics is the elaboration of techniques for discovering and classing
linguistic elements, and that grammars are inventories of these elements
and classes. Instead, he saw grammar as a theory of the sentences of the
language; and he saw the major methodological problem as the construc-
tion of a general theory of linguistic structure in which the properties of
grammars (and of the structural descriptions and linguistic levels derived
from them) are studied in an abstract way (Chomsky 1964b).
Following his presentation Chomsky engaged in debate from the floor
with leading structuralists of the past generation, whose views he was
opposing. While some of them had hoped to defeat the young upstart once
194
Linguistics: The Search for Autonomy
and for all, there was a very different outcome. As Newmeyer recounts on
the basis of the transcripts, "Here we can see linguistic history documented
as nowhere else-Chomsky, the enfant terrible, taking on some of the giants
of the field and making them look like rather confused students in a
beginning linguistics course" (1980, p. 35). In the years that followed, this
scene was to be repeated countless times, as Chomsky's steadily growing
body of admirers and followers did battle with the leaders of the once-
dominant structuralist school and demonstrated with numbing regularity
the inadequacy of earlier approaches and the promise of the new genera-
tive approach to grammar.
Chomsky's approach was indeed revolutionary-cataclysmic in rela-
tion to earlier linguistics; iconoclastic in relation to the prevailing empiri-
cist, inductivist, and nominalist temperament. Just as scholars like Levi-
Strauss in anthropology, Bruner and Miller in psychology, and Quine and
Putnam in philosophy had challenged the dogmas of their chosen disci-
pline, so Chomsky confronted received opinions in the linguistics world.
And because his approach was exceedingly formal, and the study of lin-
guistics relatively advanced in the human sciences, the scope of his claims
stood out with stark clarity.
Moreover, and this is part of the intellectual history of the time as
well, Chomsky's talents as a polemicist and his penchant for debate were
perhaps unmatched in recent scientific history. He liked to debate, was
brilliant at it, and was willing to argue tirelessly with all those interested
in his topic. He also inspired many students and colleagues to join him in
what seemed like a crusade-a crusade against the reflexive (and unreflec-
tive) empiricism of the previous generation, a crusade in favor of his views
of syntax, of language, of innate knowledge, of cognitive science, and, even
in many cases, of the contemporary political scene. In the end, many of
Chomsky's own supporters abandoned him, and today his particular lin-
guistic views are apparently held by only a minority of workers in syntax.
Yet even his most severe critics would not be studying what they are
studying, in the way that they are studying it, without his indispensable
and awesome example. And, indeed, both in the power of what he said and
in the uncompromising way he said it, Chomsky's work made a deep
impression from the first on nearly everyone who took the trouble to
study it.
All the same, as the excitement of those early days of transformational
grammar receded somewhat from view, it is easier to discern continuities
with earlier times. At the beginning of this chapter, I noted that Edward
Sapir, two generations older than Chomsky, was already dealing on an
intuitive level with many of the problems to which Syntactic Structures was
explicitly addressed. Going back much farther, the general epistemological
195
II I THE CoGNITivE SciENcEs : HISTORICAL PERSPECTIVE
approach that influenced Chomsky was taken from the work of Descartes,
and even more from those "Cartesian linguists" who were struck by the
creative powers of human language and who (like Wilhelm von Hum-
boldt) searched for ways to capture its regularities. And closer to home,
Chomsky shared the belief of his structuralist teachers in the importance
of studying syntax or grammar separately; the need (stressed by Roman
Jakobson and Louis Hjelmslev [1953]) for an essentially mathematical or
formal approach to language; the link (noted by Leonard Bloomfield) be-
tween the problem of describing a language and the problem of accounting
for language acquisition; and the conviction (with Ferdinand de Saussure)
that the study of langue (or competence) differs profoundly from the study
of parole (or performance). What Chomsky did, however, was to put these
various strands together: he showed that one could harness a formal-
logical analysis to illuminate intuitive relations among sentences, in service
of the goal of generating a system of rules which could account for all, but
only all, the correct sentences in the language. As Freud said of his discov-
ery of the unconscious through the "royal road" of dreams, such insight
falls to one's lot but once in a lifetime.
As far back as the classical era, scholars were arguing about the regularities
of language, and its ultimate characterization: is language part of nature or
part of culture? A system of grammatical categories, based on Greek and
Latin, evolved, and the study of rhetoric flourished as never since. Nor was
this interest exclusive to the Western world; in India in the fourth century
B.c., for example, Panini devised a grammar that was more diverse and in
some respects superior to that proposed by Greco-Roman scholars. During
the Renaissance, the tradition of classical philology reached great heights,
when it was applied to textual and critical analysis in various fields. As
scholars became aware of other languages, through travel and colonizing,
empirical knowledge accrued concerning the various languages of Asia,
Africa, Oceania, and the New World. There were even scientific expedi-
tions just to collect information about exotic languages (Robins 1967).
An important event in the history of linguistics occurred in 1786 when
Sir William Jones, an English orientalist, reported to the Bengal-Asiatic
Society that Sanskrit bore a striking resemblance to the languages of Greek
and Latin. Jones put forth the provocative hypothesis that all three lan-
guages must have come forth from a common source, which perhaps no
196
Linguistics: The Search for Autonomy
longer exists. This statement jostled the scholarly world, which had hith-
erto assumed that Latin had been a sort of corrupt Greek, and that most
European languages had just been derived from Latin. But a similarly
offhand explanation of the relationship between ancient languages and the
newly examined Sanskrit was not possible.
Jones's discovery ushered in a new conception of linguistic history,
linguistic phylogeny, and the processes of linguistic change. Earlier it had
been assumed that all languages had developed in the few thousand years
since Creation, and that Hebrew was the original language. Now instead,
linguists and philosophers became caught up with discovering the laws
that governed changes in language. Working in the manner of taxonomic
biologists, the linguists set up various language families (Semitic, Algon-
quin, Finno-Ugrian, and so on) and developed a theory of language
changes. Jakob Grimm found systematic correspondences between the
sounds of German and those of Greek, Latin, and Sanskrit. For example,
he noted that where Gothic (the oldest surviving Germanic language) had
an f, Latin, Greek, and Sanskrit frequently had a p. To account for these
relationships, he postulated a cyclical "soundshift" in the prehistory of
Germanic, in which the original voiceless (or unaspirated) stops became
"aspirates" (thus p becomes 1). But the systematization was not complete:
Grimm and his associates were willing to countenance exceptions to such
rules (Bolinger 1968).
The Neo-Grammarians
In the 1860s and 1870s, in a manner paralleling the Chomskian circle
a century later, a group of young rebels called the ]unggrammati/cer (or
neo-grammarians) attempted to put this situation in order. Not satisfied
with a mere collection or taxonomy of regularities, of the sort that Grimm
and his associates had noticed, the neo-grammarians claimed that the laws
of phonemic change admit no exceptions. Hermann Paul asserted in 1879,
"Every phonemic law operates with absolute necessity: it as little admits
of an exception as a chemical or physical law" (quoted in Brew 1968, p.
177}. Any apparent exceptions to these laws must be explained by refer-
ence to other laws, and there was little tolerance of theory for theory's
sake. As the neo-grammarians H. Osthoff and K. Brugmann declared:
197
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
198
Linguistics: The Search for Autonomy
just as any piece can stand for a bishop, any sound can stand for the
concept apple: all that is required is conventional agreement. Second, no
entity by itself has meaning: an entity accrues meaning only in terms of
its relationship to the other entities in an ensemble. Thus, the meaning of
a rook is only achieved in relation to what the other pieces can do in
principle and to what position they occupy at the present time; by the same
token, the meaning of a word (or a part of speech) is significant only in
terms of the other words in the language, and of the particular words with
which that given word happens to be concatenated at a specific moment
in time. Just as any move in a game of chess influences the complexion of
the game, and alters all the relations, so, too, any introduction of a new
word has reverberations across all other words in the utterance, text, or
entire language. And so language must be seen as an organized totality,
whose various parts are interdependent and derive their significance from
the system as a whole.
Saussure's focus on language as a system, and his interest in describing
that system in synchronic terms, were probably his most lasting contribu-
tions to the emerging science of language. (The Gestalt psychologists were,
at the same time, making similar points about the interrelationships among
visual perceptual elements.) However, other expeditions which he made
into the region of language were also telling. Saussure introduced a distinc-
tion between language (or langue) as a total system of regularities (like the
rules governing every game of chess) and speech (or parole) which are the
actual utterances themselves (like the moves of a particular game). While
parole constitutes the immediately accessible data, the linguist's proper
object is the langue of each community: the regularities in the lexicon,
grammar, and phonology which each person imbibes by being reared in a
particular speech community. Saussure shrewdly pointed out that a parrot
may speak the words but does not know the system of language; whereas
a person who knows the language of English may choose not to speak it.
(Shades of Searle's Chinese room conundrum!)
In some ways Saussure anticipated linguistic theory of today, and
several of his distinctions and conceptualizations are now part of the
working assumptions of every practicing linguist. For instance, he viewed
language as a cognitive system contained within the head of the individual
speaker-a mentalistic perspective that came perhaps readily to a French-
speaking scholar. But it is important to note differences between his out-
look and that of current workers. For one thing, Saussure was wedded to
a description or inventory of the elements in language and to building up
from the simplest to the most complex elements. This atomistic orientation
stands in sharp contrast to the Chomskian goal of a system of generative
rules which will yield only acceptable utterances. Moreover, Saussure paid
little attention to the properties of a sentence; in contrast, in Chomsky's
199
II I THE CoGNITIVE SciENCES : HisTORICAL PERSPECTIVE
200
Linguistics: The Search for Autonomy
in terms of such perceptual, physical, and motor properties. These distinc-
tive features become aligned into simultaneous bundles of sounds which
are the phonemes of language. While one might think that there is a huge
multiplicity of features, Jakobson and his colleague Morris Halle reported
-again, in the significant year 1956-a mere twelve basic oppositions out
of which each language makes its own selections.
In its treatment of phonology along these lines, the Prague school
made its most characteristic and enduring contribution. Note that this
approach, while superficially similar to that of the descriptive structural-
ists, was actually motivated by different concerns. The Prague scholars
were interested in explanation, not just description, and had no hesitation
in imputing psychological reality to their linguistic accounts. Rather than
viewing the features as simply "convenient" ways of describing sounds,
members of the Prague school believed that they possessed physical and
psychological reality: the nervous system had evolved so as to ensure the
proper production and discrimination of these features. Features can be
defined independently of their appearance in any particular language; and
when found, they are accounted for by their place in a general theory of
phonology. Theory guides the discovery of phenomena rather than the
other way around.
The phoneme marked an important advance. First, it turned out that
the phonemic system of every language could be characterized in terms of
a small number of binary feature oppositions. Second, the features allowed
the formulation of generalizations that could not be stated in other struc-
turalist models. Also, the positing of features made possible the develop-
ment of an evaluation procedure: the optimal analysis should need the
least number of feature specifications per phoneme. These factors were
soon to exert influence on linguists who sought to bring comparable rigor
to the still fledgling study of syntax.
By no means did the Prague school restrict itself to phonology. Indeed,
Roman Jakobson and his colleagues were driven by a lust to write about
nearly every aspect of language, from phonology to poetry, and about a
large band of languages as well. Invading the field of psychology, Jakobson
(1941) sought to apply his interest in sound laws to a wide variety of
populations. He discerned a sequence of sound development in normal
children, one that follows what he called the law of maximal confrasf; and he
went on to demonstrate that this developmental progression is reversed in
an orderly way as a result of damage to the human brain. Jakobson
showed, furthermore, that the same kinds of contrast that had worked for
the description of phonology could also be applied to choices made at other
linguistic levels, ranging from syntax to pragmatics. And in a fertile metho-
dological contribution, Jakobson pointed out that most linguistic contrasts
are not equivalent or reversible: one pole of the contrast is more basic (or
201
II I THE CoGNITIVE SciENCES : HisTORICAL PERSPECTIVE
202
Linguistics: The Search for Autonomy
linguist, Leonard Bloomfield, took a somewhat different historical tack
(Hockett 1968a ). From a contemporary cognitive perspective, Bloomfield
made two contributions-one wholly beneficent, the other far less so. His
positive work involved the development of methods and notations for the
study of unfamiliar languages. As a young instructor, Bloomfield began to
study various exotic languages-for example, Tagalog, the principal In-
donesian language of the Philippines. Bloomfield worked from dictation,
taking down an extended discourse, which he then subjected to detailed
analysis. Rejecting the model of Latin grammar, which had guided many
of his predecessors, Bloomfield recorded the sound patterns of each partic-
ular language, noting both its similarities and its departures from other
classical languages. From the study of Tagalog, Bloomfield moved to Ma-
laya-Polynesian and Algonquin. In each case he developed appropriate
methods for recording the language and struggled to find the best way of
describing its sound patterns and grammatical regularities. The identifica-
tion of constituent structures, as a means of syntactic analysis, is an
outgrowth of Bloomfield's work.
In mastering the properties of these various languages, Bloomfield was
not content simply to pursue his own research. He came to see the need
for a cadre of trained linguists. Soon he founded a linguistics society and
launched a series of publications. He saw linguistics as a natural science
-one patterned after physics:
As the physicist need not follow the path of each particle, but observes their
resultant action in the mass, and their individual actions only when these in tum
group themselves into a deflection of the mass condition (as in radio-active sub-
stances) and rarely has occasion to watch the impingement of a single particle, so
in linguistics we rarely attend to the single utterance or speaker, but attend to the
deviations of utterances and speakers only when they mass themselves into a
deflection of the total activity. (Bloomfield 1925, p. 2)
203
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
Human beings do not live in the object world alone, nor alone in the world
of social activity as ordinarily understood, but are very much at the mercy of the
particular language which has become the medium of expression for their society.
It is quite an illusion to imagine that one adjusts to reality essentially without the
use of language and that language is merely an incidental means of solving specific
problems of communication or reflection. The fact of the matter is that the "real
world" is to a large extent unconsciously built up on the language habits of the
204
Linguistics: The Search for Autonomy
group .... We see and hear and otherwise experience very largely as we do because
the language habits of our community predispose certain choices of interpretation.
(1921, p. 75)
In the hands of his student Benjamin Lee Wharf, Sapir's ideas were
further developed. Wharf did not shrink from the conclusion that even our
most basic notions might be derived from language. While Newtonians
might think that space, time, and matter reflect the same intuitions every-
where, Wharf demurred. He claimed that the Hopi Indians have totally
different conceptions of these Kantian categories, and that these distinc-
tions have grown out of the different ways in which their language parses
the universe. As he put it:
We dissect nature along lines laid down by our native languages .... The
world is presented in a kaleidoscopic flux of impressions which has to be organized
by our minds-and this means largely by the linguistic systems in our minds.
. . . No individual is free to describe nature with absolute impartiality but is
constrained to certain modes of interpretation even while he thinks himself most
free. (1956, pp. 153, 213-14)
Many linguists were intrigued by the work of Wharf and Sapir, even
as many psychologists and anthropologists found it an excellent source of
fertile hypotheses. (In chapter 12, I shall examine some of the cognitive
scientific work it inspired.) And yet, like much of the work in semantics
of the period, it was considered peripheral to the mainstream of linguists
-too difficult to lay one's hand on, too much contaminated by the "real
world" and by the dangerous notion of "meaning," not susceptible to
sufficient documentation or proof. Indeed, Sapir and Bloomfield stood in
marked contrast to one another. Bloomfield assumed the pose of a rigor-
ously scientific worker, concentrating on methodology and formal analy-
sis. His insistence on a strictly mechanistic approach to meanings and his
pessimistic attitude toward semantics contributed to the relative neglect of
that subject matter. Far more reminiscent of Jakobson, Sapir ranged widely
and was not afraid to touch on culture, conceptualization, and the world
of meaning. Paradoxically, he shared many of Chomsky's intuitions about
linguistic structure but little of the formalistic inclinations of either Bloom-
field or Chomsky. A union of Bloomfield-Sapir approaches and insights
has yet to take place in linguistics.
205
II I THE CoGNITIVE SciENCES : HisToRICAL PERSPECTIVE
Period Pieces
It is not possible in a survey to detail the particular models, with their
numerous alterations, put forth by Chomsky and his followers in the years
following the publication of Syntactic Structures. Nonetheless, it is important
to stress that Chomsky's contribution did not occur at a single moment in
time and does not consist of a single body of theory. Indeed, few of the
specific notions put forth in Syntactic Structures have remained in their origi-
nal form (even though the overall program has been maintained with
considerable fidelity). Kernel sentences were soon dropped; the number of
transformations was reduced and simplified enormously; the relation
among the levels of syntax, semantics, and phonology has come to be
understood along somewhat different lines. Certain topics introduced in
that initial volume have been abandoned; while others, merely hinted at
in 1957, have spawned entire research programs.
The first major new statement came in Chomsky's monograph Aspects
of the Theory of Syntax (1965). This book presented what is now termed the
11
Standard Theory," the theory that for many years was the staple of
introductory textbooks in linguistics. In the Standard Theory, there are no
longer initial kernel sentences. Instead, one now starts with the base gram-
mar, which generates an initial phrase marker, the deep structure. Major
operations are performed upon this deep structure. There is a transforma-
207
II I THE CoGNITIVE SciENCEs : HISTORICAL PERSPECTIVE
tional component which converts the initial deep structure into other
structures, the final of which is the surface structure; in this theory, most of
the transformations are obligatory.
Two interpretive components operate on these core syntactic compo-
nents. Deep-structure relations are interpreted by a semantic component:
thus, the information necessary for semantic analysis must be represented
in the deep structure. Phonological interpretation occurs on the surface-
structure string: thus, the initial deep syntactic structure is "read" for
meaning, while the ultimate surface structure is "read" for sound. In the
1965 version, both the base and the transformation rules were uncon-
strained, permitting a large variety of possible base and transformational
systems.
The Standard Theory was a more ambitious theory, attempting in part
to accomplish for semantics what had been modeled for syntax alone in
Syntactic Structures. It also proved far more controversial and, because of
various insufficiencies, eventually had to be abandoned. Particularly con-
troversial was the notion of deep structure. Deep structure is a technical term,
having nothing to do with profundity. Nonetheless, it was frequently
misinterpreted in this fashion. Moreover, even among those who under-
stood its technical scope, deep structure was seen as "too deep" by some
critics and as "not deep enough" by others. The move of making semantics
an interpretation of core syntactic arrangements was also widely debated.
We can distinguish two broad lines of attack on Chomsky. One line
of attack came from the conservative branch of linguistics-from scholars
still attached to the structuralist perspective of the Bloomfield days.
Charles F. Hockett selected himself to answer Chomsky on behalf of the
traditional interest groups in structural linguistics. In The State of the Art
(1968b ), Hockett argued that the fundamental program of transformational
linguistics was flawed. Chomsky had bet on the wrong scientific model. He
had compared language to a formal discipline like logic or mathematics;
whereas it is an empirical science like chemistry. Only humanly invented
systems like logic have the regularity Chomsky sought in language; in
contrast, natural language is ill defined, and the theories of computability
and algebraic grammar he sought to apply are irrelevant. "This knocks the
props out from under current mathematical linguistics, at least in the form
of algebraic grammar, whose basic assumption is that language can be
viewed as a well-defined subset of the set of all finite strings over a
well-defined finite alphabet" (1968b, p. 61). Hockett also rejected the
separability of grammar and semantics (which he had earlier espoused) and
the separation of both from the rest of culture. Autonomy within linguis-
tics, and from other disciplines, was abandoned. In his view, the "grammar
of language" in Chomsky's sense simply does not exist.
208
Linguistics: The Search for Autonomy
What is striking was the absence of response to this monograph.
Except for a dismissive review by Chomsky's one-time associate George
Lakoff (1969), this work of criticism was ignored. Apparently by the late
1960s, ripostes from the earlier generation had simply become irrelevant.
In accordance with Thomas Kuhn's (1970) description of paradigm changes
in the sciences, the Chomsky point of view took over, not by convincing
the previous generation that it had been in error, but by winning the
allegiance of the most gifted students of the succeeding generation.
Some members of the younger generation felt, however, that
Chomsky was not sufficiently radical. In the late 1960s a major attack on
his work was launched by certain linguists who had been sympathetic to
his general position. Led by Chomsky's former students, George Lakoff
and}. R. Ross (1976), these critics questioned the simple positing of two
levels of analysis-a deep structure, to be interpreted semantically, and a
surface structure to be interpreted phonetically-and also called into
doubt the autonomy of syntax from semantics. Ultimately, these critics
abandoned simple deep structure in favor of grammars whose underlying
structures were much deeper and closer to semantic representation them-
selves. (Another group of critics, which eventually included Chomsky
himself, embraced grammars in which the underlying structures were
shallower and closer to surface structure than had previously been en-
visaged, and were equipped with much richer lexical components than
had earlier been the case.) The generative semanticists contended that
there was no dear-cut distinction between syntactic and semantic rules,
and that a level of syntactic deep structure defined as the initial genera-
ting component could not be sustained. Instead, they set up rules that
take semantic representations as their input and yield surface structures
as their output, with no intervening level of deep structure. In contrast,
the interpretive semanticists, including Chomsky, transferred more of the
work of syntax into the semantic component, so that deep structure has
gradually moved closer to the surface structure of a sentence (Smith and
Wilson 1979).
The debates between the generative semanticists and the traditional
Chomskians, and, eventually, between the generative semanticists and the
interpretive semanticists, were ardent and often vicious. At stake were
particular claims about how best to represent word meanings, as well as
competing views about how to construct linguistic theories. Names were
called and epithets were hurled in a manner that shocked even seasoned
polemicists. When all the hue and cry had died down, however, the gener-
ative semanticist school lost its steam, and most of its principal adherents
abandoned the field of syntax altogether. Meanwhile, in his persistent
way, Chomsky has continued to investigate syntax. It is true, however,
209
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
that the problems pointed out by the generative semanticists have stimu-
lated the Chomsky contingent to develop certain notational forms (using
symbolic logic) as well as a plethora of new and highly abstract mech-
anisms, all designed to deal with inadequacies in the traditional transfor-
mational approach. Perhaps most decisively, generative semantics helped
to sound the death knell for the transformational component; this once
central component has been entirely eliminated on some accounts or radi-
cally simplified on others.
Since the late 1960s, Chomsky's brand of generative grammar (as it came
to be called) has undergone a series of changes. It is not possible to intro-
duce and follow the intricacies of the extended standard theory or of its
current replacement government and binding theory, let alone the specifics
of trace theory or X-bar theory-the major landmarks in the development
of his approach up to the present. I can, however, indicate some of the
broad trends in the theory. Over the years, Chomsky has steadily nar-
rowed the definition of his object of study. Never interested in language
as an overall communication system, he now questions whether language
per se is a system worth trying to study at all. Viewing language as a more
abstract notion than grammar, more remote from actual mechanisms,
Chomsky is more firmly convinced than ever that linguists should concen-
trate on solving the issues of syntax (1982, p. 14). There are no more
attempts to systematize semantics.
Responding to many criticisms, Chomsky has attempted to reduce the
expressive power of transformational structures. It turns out that the origi-
nal grammars had been so powerful that they could not illuminate the
crucial issue of what is permissible in a human language (see Peters and
Ritchie 1973). If one wanted to know which grammars might characterize
human performance, it was necessary to restrict what a candidate grammar
could do. This attempt to weaken transformational power has been part
of what Chomsky has called the second major shift in linguistics-the
search for principles that would constrain the set of syntactic rules of
natural language.
Chomsky has tried to reduce the class of transformations by discover-
ing general conditions that rules must meet: he has sought to eliminate the
possibility of compounding elementary operations to form more complex
transformational rules. As he now puts it:
[M]uch effort has been devoted to showing that the class of possible transfor-
mations can be substantially reduced without loss of descriptive power through the
discovery of quite general conditions that all such rules and the representations
they operate on and form must meet .... [The] transformational rules, at least for
a substantial core grammar, can be reduced to the single rule, 'Move alpha" (that
is, "move any category anywhere"). (Mehler, Walker, and Garrett 1982, p. 21)
210
Linguistics: The Search for Autonomy
It may even be possible to eliminate the phrase-structure rules wholly
or in large part. As transformational rules and phrase-structure rules be-
come less prominent, more attention is being paid to the lexicon-to the
particular rules governing specific words. The lexicon now contains much
of the information that used to be part of the transformational apparatus.
And the notion of surface structure (now sometimes reconceptualized as
5-strucfure) becomes much richer. In fact Chomsky has even commented
that his new theory could be considered as a "one level theory," consisting
merely of base generation of S-structures (see comment in Longuet-
Higgins et al. 1981, p. 279).
How then, to characterize Chomsky's current approach to language
or, more strictly speaking, to grammar? He describes his pursuit in theo-
retical terms as a search for universal grammar (or U.G.). U.G. is genetically
determined at its initial state (in the organism) and is specified, shar-
pened, articulated, and refined under conditions set by experience, to
yield the particular grammars found in specific groups of individuals (1980,
p. 234). A theory of universal grammar is said to be an explanatory the-
ory. To know a language is to be in a certain state of mind/brain: this
state is described by a core grammar which consists of certain principles
of universal grammar, which are to be discovered by linguists. Thus,
Chomsky takes a purely realistic stance: language knowledge is a series
of states in the brain.
What are these principles to which Chomsky refers? Here we come
to what he considers to be another major shift in his theory. Until fairly
recently, his goal had been to describe the various rules that individuals
must somehow know if they are to know a language. But now he has come
to the conclusion that it is more productive to speak of various princi-
ples that govern language use. These principles begin in human biology.
They determine the kinds of grammar that are available in principle
(Chomsky 1979). The principles of U.G. are various subsystems, which go
by names like binding theory, control theory, government theory, theta theory, and the
like-each of which features a limited degree of parametric variation.
If U.G. is sufficiently rich, then even limited linguistic data in the
environment should suffice for developing rich and complex linguistic
systems in the mind. There is no need to talk of the acquisition of rules.
Rather, each of the systems of U.G. has certain parameters associated with
it, and these are set or fixed in light of the data encountered by a person
(ordinarily a young child) in the course of acquiring his native language.
Slight changes in the values of the parameters proliferate throughout the
system to yield what on the surface may be rather different language
structures (Chomsky 1982}. The grammar of a particular language that
ultimately emerges can be regarded as a set of values for each of these
211
I I 1 THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
parameters: the overall set of rules, principles, and the like constitute U.G.,
part of the human language faculty.
We see here the union of two visions in the science of the mind. One
vision, stemming from philosophy and dating back to Plato, features lan-
guage as a kind of idealized object, governed by a small set of universal
principles, having relatively few parameters (Chomsky 1981). The other
vision, coming from biology, puts forth language as an organic system, or
module, which has the potential to develop in a small and delimited range
of ways; the particular path of development yielding a "core grammar" is
determined by the kinds of information encountered by the organism.
Embracing these two visions, one can account at once both for the similari-
ties across all languages (thanks to universal grammar) and the distinct
differences among particular languages (thanks to the variations in param-
eter setting).
What is one to make of such a complex theory which has undergone
critical changes in the course of a few years and will doubtless change
further? One approach, sometimes adopted by Chomsky's former stu-
dents, is to stress the differences in the theory from one period to another
and to attribute these differences to efforts to shore it up in the light of
competing accounts. Such a critical point of view has been put forth by
George Lakoff (1980). In the view of this former Chomsky student,
Chomsky had to eliminate deep structure and transformations because it
had proved impossible to maintain the principal assumption of modularity
-the independence of syntax. Once Lakoff and his associates had shown
that meaning and use affect virtually every rule of syntax, Chomsky had
to narrow progressively the domain of syntax: rules once regarded as
clearly within the domain of syntax were redefined as part of semantics.
On this account, Chomsky has conceded the validity of criticisms in prac-
tice while, in his explicit remarks, denying that he has done so. The
historian of American linguistics, Frederic Newmeyer, offers another inter-
pretation of this phenomenon. In his view each generation of Chomsky
students continues to work on the major ideas proposed by their teacher
during their own intellectually formative years:
212
Linguistics: The Search for Autonomy
have some general sympathy with generative grammar. However, he
does not like to concede this influence explicitly. He is not easy to dis-
pute with, and does not suffer criticism gladly, and so tends to be dismis-
sive in writing even (or perhaps especially) of those scholars whose work
has had some influence on him. This rhetorical style tends to polarize
discussion.
Despite undeniable shifts in emphasis and strategy, however, it is
remarkable how Chomsky has adhered to the program he initially pre-
sented to the scholarly community in Syntactic Structures and had been pur-
suing from his early twenties. The centrality of syntax, the belief in a
transformational component, the view of semantics as an interpretation of
basic syntactic relations-all these have endured.
Chomsky is correct in saying:
My major interest has been to make precise the basic principles which enter
into the knowledge of language that has been attained by the speaker-hearer; and
beyond that, to try to discover the general theoretical principles which account for
the fact that this system of knowledge, rather than something else, develops in the
mind when a person is placed in a certain linguistic environment. In a general way,
I might say that I am still working very much within the framework of ... early
unpublished work. (1979, p. 113)
Any growing science will be constantly changing. Chomsky has been one
of the principal revisers of his own theory, sometimes in radical directions,
even though the same vision has guided him from the start and he has been
relentless in pursuing it.
As he moves to increasingly abstract characterizations, involving tech-
nical argumentation, he has indeed lost adherents. This does not seem to
bother Chomsky, who (like many another revolutionary figure) has always
seen himself as somewhat of a loner. As he declared in a recent interview:
The particular domain into which I put most of my energies, the structure of
language, seems to me to have been a very exciting one just in the last seven or
eight years. I don't pretend to speak for any consensus in the field here, in fact,
I'm in a very small minority in the field in this respect, but I believe it's been
possible in the past few years to develop a theory of languages with a degree of
deductive structure that provides a kind of unification and explanatory power
going well beyond anything that would have been imagined even a decade ago.
Again, I don't think many linguists would agree with me about this .... I suppose
I'm in a very small minority in the field today. But then, that has always been the
case. With regard to me, it doesn't seem very different now from what it was ten
or twenty years ago. But my own views are not what they were then, and I hope
they will not be the same ten years from now. Any person who hopes to be part
of an active growing field will take that for granted. (Quoted in Rieber 1983, pp.
62-63)
213
II I THE CoGNITIVE SciENCES : HISToRICAL PERSPECTIVE
214
Linguistics: The Search for Autonomy
characterizing the data collected; at other times, the data have been used to
test the "psychological reality" of models-the extent to which linguistic
behavior unfolds according to the principles put forth by Chomsky.
After two decades of psycholinguistic work, most psychologists have
despaired of applying Chomsky's work in any direct way to their own
research. Not only are Chomsky's formulations highly abstract and subject
to frequent change (typically in a more abstract direction); but when they
have been applied, the results have not been consistent with his models,
at least in any straightforward way. For example, any number of develop-
mental psycholinguists have sought to account for language acquisition in
terms of Chomsky's categories and derivations, but these efforts have
generally been judged failures. Recently there have been interesting at-
tempts to develop learnabilify theory-a formal account of the constraints
that must be built into a child's cognitive apparatus if one is to learn
language from the data to which one is exposed (Wexler 1982; Wexler and
Culicover 1980). Steven Pinker (1984) has attempted to interpret the data
of language acquisition in terms of principles of leamability. It is too early
to say whether this approach will signal a new use for a Chomsky-inspired
(or "generative") approach within psycholinguistics or whether a Chom-
skian perspective will continue to be a minority taste.
For the most part, then, Chomsky has been more influential in psy-
cholinguistics because of the kinds of question to which he has drawn
attention than because of any direct utility of his theory for experimenta-
tion. Sometimes he and his followers have discounted empirical research
in psycholinguistics, with the disclaimer that their theories have to do with
idealized competence, and not with the facts of individual performance. To
my mind, this is an unjustified maneuver. Chomsky and his followers are
only too happy to cite empirical data when it appears to accord with their
theory. However, it is often far from clear just how directly Chomskian
ideas are meant to be translated into empirical work; and, in that sense at
least, the Chomskian reservation is justified.
Chomsky's new ideas were introduced to the philosophical world
primarily through the work of two young Princeton graduates, Jerrold Katz
and Jerry Fodor, who were teaching at Massachusetts Institute of Technol-
ogy in the early 1960s. Collaborators for several years, Katz and Fodor
(1963) also developed a model of semantics which became incorporated
into the "standard version" of transformational grammar. As has been the
case in psychology, Chomsky's ideas about language have persuaded a
minority of younger philosophers, but many others are decidedly ambiva-
lent about his specific claims. The formalism itself presents less of a prob-
lem, but a number of the core ideas have had a checkered history in
philosophy. Philosophers have reacted coolly to Chomsky's promotion of
seemingly discredited rationalist notions and to his enthusiasm for innate
215
I I I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
ideas. His ready use of terms like rules, s!rudures, systems, with (apparent)
disregard for the nontrivial technical problems involved in such concepts,
and his facile reinterpretation of leading philosophical figures of the past
(Descartes as a hero, the empiricists as villains) have proven difficult for
most philosophers to swallow. Also Chomsky's lack of interest in seman-
tics has troubled many philosophers, who find in the work of semanticist
Richard Montague some of the same formal elegance others have admired
in Chomsky's syntactic discussions (Thomason 1974).
Yet it is fair to say that, as in psychology, topics of discussion have
been materially affected by the fact that Chomsky has spoken and written.
Whether the impact of Chomsky's notions will be greater in a half-century
than they are now, or whether they will be seen as a curious aberration
within the general triumph of empirically oriented and anti-mentalistic
sciences is too early to say. Except in the field of linguistics itself, it remains
uncertain whether Chomsky's ideas will emerge as germinal and essential.
And what of other cognitive sciences? While his approach arose from
many of the same roots as artificial intelligence, several of Chomsky's main
ideas are not readily implemented in computational formats. For example,
there is no guarantee in principle that one can parse sentences using trans-
formational grammatical approaches. Moreover, A.I. is very much oriented
toward practical problems of designing programs that understand sen-
tences or stories, and Chomsky's syntax-centered framework is not suited
for the main issues of understanding discourse. Accordingly, computer
scientists like Roger Schank have been publicly hostile to the theory,
taking the position that semantics and pragmatics are central in language
and that syntax is relatively unimportant. Schank has also attacked the
modular notion: "It is impossible to produce a model of language alone
... apart from beliefs, goals, points of view and world knowledge" (1980,
p. 36). Other scholars, like Terry Winograd, have borrowed some ideas
from transformational grammar but have not made Chomskian theory
pivotal to their systems. Efforts to parse sentences using transformational
ideas have been few thus far (though see Berwick and Weinberg 1983;
Marcus 1980). Even through Chomsky's formal elegance has appealed to
A. I. researchers, his Platonic view is even more remote from most computer
scientists than from the average psychologist and philosopher. For his part,
Chomsky has been rather critical of research in artificial intelligence,
finding it mostly unmotivated and ad hoc: he does, however, admire David
Marr's work on vision (1982).
216
Linguistics: The Search for Autonomy
have sometimes been attracted to the artificial-intelligence perspective.
One such researcher is Joan Bresnan, formerly a colleague of Chomsky's
at M.I.T., who has concentrated on developing a theory of language which
is psychologically real (1978, 1981). In opposition to Chomsky, who (as I
noted) pays little attention to how his derivations might be realized by an
11
individual speaker operating under real-world" constraints, Bresnan and
her colleagues have fashioned a perspective designed to illuminate how an
individual will perceive or produce .language. In her lexical-functional the-
ory, there is no transformational component. The information traditionally
embedded in the syntactic components is now placed in the individual's
lexicon-one's knowledge of specific words. Lexical-functional grammar
provides each sentence with two structures: a constituent structure, which is
similar to a surface structure (or phrase-marker tree) in the standard theory
of Chomsky; and a functional structure, which includes all the grammatical
relations relevant to the semantic interpretation of the sentence. The func-
tional structure is generated by annotated phrase-structure rules work-
ing in conjunction with lexical entries for the various morphemes in a
sentence.
In recent years, Bresnan has begun to work closely with colleagues in
the area of artificial intelligence and in psycholinguistics. The purpose of
this collaboration is to determine whether the modifications she has intro-
duced into standard syntactic theory make her position "psychologically
viable" in a way Chomsky's has never been. The results thus far suggest
that the kinds of parsing mechanisms devised in the light of her theory do
comport better with experimental data on language processing and with
the models of understanding being developed by workers in computa-
tional linguistics. For similar reasons, in his work on learnability theory,
Steven Pinker (1984) has embraced lexical-functional grammar, which he
views as a "central tendency" (or modal position) among contemporary
linguistic theories. The work of Bresnan and her colleagues has also be-
come central in the new, Stanford-based Center for the Study of Language
and Information, in which the avowed goal is to achieve an integrated
theory of human linguistic competence. The Center has just been
launched, and it is too early to say whether its dream of uniting the
philosophical, psychological, and computational aspects of language will
be met; but all of us involved in the cognitive science movement will be
monitoring progress there with abiding interest.
Much of the appeal of Bresnan's theory also accrues to another point
of view which has recently gained adherents-the theory of Gerald Gazdar
of the University of Sussex. Reverting to the generalized phrase-structure
grammars that Chomsky strongly attacked in his early publications, Gaz-
dar (1981) argues that one does not need transformations, and that even
unusual surface structures can be stated in a straightforward way; more-
217
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
When one realizes that the syntactic theory that Chomsky has been develop-
ing over the last ten years has embraced phrase structure rules, complex symbols,
a level of S-structure, a level of D-structure, a level of "Logical Form," filters,
transformations, interpretive rules, stylistic rules, coindexing conventions, and
abstract cases, among other things, it is a little surprising to hear him castigating
as "needlessly complex" an alternative syntactic theory that employs only phrase
structure rules, complex symbols, and a level of surface structure. (Quoted in
Longuet-Higgins, Lyons, and Broadbent 1981, p. 281)
A Tentative Evaluation
218
Linguistics: The Search /or Autonomy
are important in linguistics continue to dominate discussions in much of
the profession: and his ways of formulating issues have influenced all
cognitive scientists, even scholars overtly inimical to his work and young
students only dimly aware of the source of their views.
If the index of importance of a scholar is the extent to which one could
be replaced within a discipline, Chomsky clearly emerges as the most
important figure in the linguistics of recent times-equal in importance to
de Saussure and to Jakobson in earlier eras and one who may ultimately
exert an even greater influence on the direction of his field. Within the
broader cognitive sciences, his contribution is more controversial and less
secure. But on the bases of his demonstrations of the mathematical preci-
sion implicit in language, his modeling of the importance of theory-driven
research, and his potent arguments for mentalism, nativism, and mod-
ularity, I would offer my opinion that he is one of the two or three most
important and least replaceable thinkers of the whole movement. Piaget
may have turned up a greater number of important phenomena; Herbert
Simon and Allen Newell may have put forth a paradigm more widely
emulated by other investigators; but no one has framed the issues of a
cognitive science with as much precision and conviction as has Chomsky.
Moreover, while not even he has fully lived up to the rigorous criteria he
has demanded of a linguistic theory, the criteria he devised in his early
works continue to be those by which subsequent linguistic theories (and
theorists) are judged. Whether his impact will be as broad in other cogni-
tive sciences may depend upon whether the model, which has proved
fertile in the area of language, proves equally useful in other areas, such
as visual perception, logical classification, or the study of consciousness.
Chomsky himself might feel of two minds about this. On the one
hand, he is possessed of a rigorous scientific conscience and would like to
see his degree of formal precision and his criteria for explanatory adequacy
invoked everywhere. On the other hand, he has long insisted that the rules
governing language may be unique to that domain, and he is loath to
endorse notions of general cognitive structures that cut across diverse
contents. And so, while the study of linguistics in which he has pioneered
may serve as a model of how one goes about investigating other fields, it
remains an open question whether any parallels of substance will emerge.
This paradoxical situation is epitomized by Chomsky's view of lin-
guistics as a part of psychology. In repeatedly giving voice to this senti-
ment, Chomsky may appear to be suggesting that the study of language
ought simply to be incorporated into a more general study of psychology
(a catholic view). Yet it seems clear to me that Chomsky would not ap-
prove of incorporating linguistics into psychology as currently practiced.
Psychology would need to be reconfigured in Chomskian terms (a far
219
I I I THE CoGNITivE SciENCES : HISTORICAL PERSPECTIVE
220
Linguistics: The Search for Autonomy
is most interested in illuminating. If one is interested in language as an
abstract system-be it the creation of some divine force or simply a pattern
emerging from the brain-then language is appropriately studied in terms
of the kinds of taxonomic and structural category favored by linguists.
(Language here becomes a "distanced" object of study, analogous to the
solar system probed by astronomers.) If, however, one is interested in
language as it participates in human intercourse, then a view of linguistics
as divorced from other disciplinary pursuits becomes less tenable.
Consistent with this latter point of view, some scholars have come to
adopt (or to readopt) a more horizontal model of cognitive science. On this
view, it is wrong to think of the subject matter of any discipline (for
example, language as the subject matter of linguistics) as privileged: in-
deed, it is more important to connect any human activity to the range of
related fields that can be investigated. Furthermore, language does not
belong to any discipline but is instead a part of every cognitive scientific
discipline, efforts to cordon it off being artificial or wrong-headed. This
was the faith that influenced scholars like Jakobson in the area of poetics,
Sapir in his studies of language and thought, and de Saussure in his
historical studies of language. Calling for a closer integration of fields, Roy
Harris, a harsh critic of Chomsky's, declares:
221
II I THE CoGNITIVE SciENcEs : HisTORICAL PERSPECTIVE
real use of language and linguistic judgements, the grammatical system being one
of these. (1979, pp. 140, 152-3)
222
8
Anthropology: Beyond
the Individual Case
Lucien Levy-Bruhl Examines the Mind of the Primitive
223
I I I THE CoGNITIVE SciENCES : HisTORICAL PERSPECTIVE
The step which I have just taken, and hope is decisive, consists, in a word,
in abandoning a badly posed problem ... even allowing for the numerous and
characteristic cases of participation of which my six volumes are full, there still exist
doubts about the explanation . . . . I started by positing a primitive mentality
different from ours ... a position which I have never been able to defend well, and
in the long run an untenable one.... The thesis thus extenuated and weakened
is no more defensible . . . . Let us entirely give up explaining participation by
something peculiar to the human mind. . . . There is not a primitive mentality
distinguishable from the other. (Quoted in Cazeneuve 1972, pp. 86-87)
224
Anthropology: Beyond the Individual Case
remote from those who doubted there was any fundamental difference
whatsoever. Levy-Bruhl took refuge in the argument that, perhaps, it was
scientific thought that is unusual: in the paraphrase of Jean Cazeneuve,
"What has been described to us under the name of primitive mentality is
undoubtedly a permanent structure of the human mind but in our society
this structure is blurred by the supremacy of scientific thought whereas it
remains in the foreground among preliterate peoples" (1972, p. 22). He
concluded with a sentiment that would gain widespread endorsement
today: "The fundamental structure of the human mind is the same every-
where" (p. 12).
The debate that raged in Levy-Bruhl's mind echoed a discussion that
has been widespread in the West ever since humans first became aware of
the existence of exotic populations and began to ponder their relation to
more familiar pockets of humanity. Though the debate extended to the
morals and values of these alien populations, keen interest always centered
on the quality of their thought processes. Did they have the same logics
as "we" do, or were there logics peculiar to each group of individuals? How
could we get inside the mind of the primitives, and discern the world in
the way they see it? Those scholars who acquired a professional interest
in these questions were "at risk" to become anthropologists.
In the time of Levy-Bruhl, early in this century, the sources of evi-
dence were principally textual: one read the myths or, less frequently,
transcripts of conversations with preliterate individuals, and then drew
conclusions about the kinds of thought reflected in them. But such her-
meneutic methods could not attain scientific status. And so, increasingly
in the twentieth century, anthropologically oriented scholars repaired to
the field to examine firsthand the thought processes of "their people." This
move to "case studies" was certainly an important step in an empirical
direction. The problem with individual fieldwork, however, was that it left
a great deal of discretion-perhaps too much-in the hands of a single
investigator or a small cadre of fieldworkers.
A significant chapter of the history of anthropology in this century
has been a search for methods of inquiry that were less idiosyncratic, more
reliable (Herskovits 1953; Kardiner and Preble 1961). Since it was generally
not practical for large teams of investigators to visit the same site-and if
investigators visited at widely disparate times, they might not be witness-
ing the "same" peoples (Freeman 1983)-a premium was placed on more
objective methods which could be employed by a single investigative
team. This need gave rise in the 1960s to the field of e!hnoscience, a seemingly
objective set of empirical measures by which one could assess the thinking
processes of peoples everywhere.
Ethnoscientists had no problem in invoking a representational level.
225
II I THE CoGNITIVE SciENCES : HisTORICAL PERSPECTIVE
226
Anthropology: Beyond the Individual Case
subsistence patterns, state of commerce, climate, and soil. In his view,
natural surrounds and conditions of climate played a more decisive role in
savage societies than in more developed societies. As an early functionalist,
he sought within each society a motivated basis for such seemingly brutal
practices as cannibalism, slavery, or idol worship (De Waal Malefijt 1974).
The advent of the Enlightenment, with its ardent belief in rationality
and equality, sharpened the debate about the mentality of the primitive.
On the one hand, to the extent that rationality was the standard, the
Western mind seemed qualitatively more advanced than that of the appar-
ently confused savage. On the other hand, to the extent that equality was
stressed, another set of conclusions seemed warranted. Such practices as
slavery, and such beliefs as the superiority of one group over another, were
seen as regressive or anachronistic. These emerging egalitarian points of
view posed difficulties for those of a religious persuasion, who needed to
explain why some groups believed in a single God, while others clung to
polytheistic notions. The notorious Bishop Whately claimed that savages
could not be helped and were best thought of as members of a different
species. Darwin posed a different threat to members of established
churches who now had to contend with his demonstrations that all hu-
mans were descended from a line of forerunners, dating back millions of
years, and that human beings could not be thought of apart from the rest
of the Natural Order (De Waal Malefijt 1974).
The stage was set for more systematic thinking about different human
groups. Various scholars helped to initiate the scientific study of society
and culture, but the person most often granted this honor (or responsibil-
ity) is an Englishman of the late nineteenth century, Edward Tylor. Work-
ing at the same time as Wilhelm Wundt, the founder of modern psychol-
ogy, Gottlob Frege, the inventor of modern logic, the neo-grammarians in
linguistics, and the first generation of experimental neuroscientists, Tylor
published his magnum opus, Primitive Culture, in 1871. Having toured
America and Mexico in the 1860s and obtained thereby a vivid notion of
cultural differences, Tylor undertook in his book a rationalist assault on
the divine inspiration of religious beliefs. According to his revisionist
perspective, human culture and religions were products of a natural, law-
governed evolution of human mental capacities.
A new field requires definitions, and Tylor produced the most-often
227
II I THE CoGNITIVE SciENcEs : HISTORICAL PERSPECTIVE
228
Anthropology: Beyond the Individual Case
certain practice came about, Tylor could now ferret out statistical relations
between different institutions or practices.
From our contemporary perspective, Tylor's views may seem to blend
with those predecessors who saw the primitive as just a pale version of
civilized modem man. In fact, however, Tylor helped to undermine this
view. To argue that culture actually exists among all men, in however
crude or primitive a form, was a major step toward a more relativistic point
of view. Moreover, Tylor's belief that all groups harbor vestiges of the
past, and that behavior can be understood and justified if seen in context,
served to bring the primitive person closer to the circle of civilized modem
man (Stocking 1982). As Abram Kardiner and Edward Preble have com-
mented, to overcome the ecclesiastics "who would create an impassable
gulf between civilized man and his primitive ancestors, Tylor had to show
that the 'rude savage' was potentially an English gentleman" (1961, p. 77).
Could Shaw's Pygmalion be far behind?
229
II I THE CoGNITIVE SciENCEs : HISTORICAL PERSPECTIVE
230
Anthropology: Beyond the Individual Case
231
II I THE CoGNITivE SciENCES : HisTORICAL PERSPECTIVE
His greatest contribution to science, and, at any rate, the one we can best
appreciate, was the development of descriptive language study. The native lan-
guages of our country had been studied by some very gifted men, but none had
succeeded in putting this study on a scientific basis .... Boas amassed a tremendous
body of observation, including much carefully recorded text, and forged, almost
single-handed, the tools of phonetic and structural description. (1943, p. 198)
232
Anthropology: Beyond the Individual Case
tions and are now available to the general population. Boas's own explora-
tions had convinced him that Indians and Eskimos could appreciate ab-
stract language and thought if confronted with it, but that this set of
concerns was simply not part of their habitual intellectual milieu. Overall,
then, Boas was one of the first, and one of the strongest, advocates of the
view that primitive and modem individuals possess essentially the same
cognitive potential; an enduring suspicion of unwarranted or undocu-
mented dichotomies between "primitive" and "modem" is one of his most
compelling contributions to contemporary anthropological thinking.
From the start of his career, Boas had been suspicious of any attempt
to evaluate one human being, or one human group, as better or worse than
another; as his career thrived, he used his increasingly powerful and in-
creasingly public platform to counteract all forms of racism. At least as
formidable as his contributions to anthropology was Boas's role in chal-
lenging the still-prevalent views in the United States that members of one
social or ethnic group were more intelligent or morally virtuous than
another. That there was no scientific basis for labeling one group as inferior
to another was the theme that Boas kept reinforcing in his writings. It is
a theme that became part of the fabric of social science from the 1930s until
the present.
Reactions to Boas
As Boas's contributions were primarily methodological, and as he had
an instinctive distrust of overarching theories, it is not surprising that his
most vocal critics in the next generation were those with a strong theoreti-
cal position to defend. Leslie White (1963) and Marvin Harris (1968),
devotees of evolutionism who were sympathetic to Marxism, portrayed
Boas as one who refused to take a stand on the relationship between one
culture and another, and who, in his passion for data about particular
individuals and groups, neglected the material and technological basis of
human activities. A. R. Radcliffe-Brown (1940, 1952), one of the leading
theorists on the British anthropological scene, stressed the importance in
anthropology of an undergirding theory; he promoted a Durkheimian
approach, in which the needs for group solidarity exert a decisive impact
on kinship structures and on the actions and beliefs of individuals. Rad-
cliffe-Brown also saw cultures as part of a social system, as "organisms"
which evolve toward increasing diversity and complexity. All of this theo-
rizing did not sit well with Boas, who distrusted generalizations about the
needs of a group or about a culture as a whole.
Another line of research was the functionalist approach of Bronislaw
Malinowski (1961, 1968): akin in some ways to behaviorism in psychology
233
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
234
Anthropology: Beyond the Individual Case
discussing the history of linguistics, Sapir and his student Benjamin Lee
Whorf came to believe that the language used by a group has been a
principal determinant of the belief structures and the ways of thinking of
that population. Indeed, Sapir came to see language as a guide to social
reality:
It powerfully conditions all our thinking about social problems and processes .
. . . It is quite an illusion to imagine that one adjusts to reality essentially without
the use of language and that language is merely an identical means of solving
specific problems of communication or reflection .... No two languages are ever
sufficiently similar to be considered as representing the same social reality. The
worlds in which different societies live are distinct worlds, not merely the same
worlds with different labels attached. (1929, p. 162)
235
II I THE CoGNITIVE SciENcEs : HISTORICAL PERSPECTivE
Levi-Strauss's Canons
This much was straight linguistic theory, of the sort put forth by
Jakobson and his associates in the Prague school and subscribed to, in more
or less faithful fashion, by others influenced by Saussure. The decisive step
taken by Levi-Strauss, and the core of his contribution to anthropology,
lay in his insistence-and his demonstration-that key aspects of culture
are best thought of as linguistic in nature and are best approached by the
methods of the structural linguist.
Appropriately for one who wanted to capture the attention of his
anthropological colleagues, Levi-Strauss began his inquiries by confront-
ing an area central to the concern of all anthropologists-the area of kin-
ship relations or kinship structures. To start with, he noted that in any
kinship system one has as primary data both the system of relations
between terms (father, son) and the system of relations between attitudes
(intimate, distant). As a test case for his notions of structural anthropology,
Levi-Strauss selected the classic problem of the avunculate: the relation-
ship where the maternal uncle represents family authority and exerts
certain rights over his nephew, and yet can maintain an informal, joking
relationship with that nephew. Levi-Strauss notes a correlation between
this set of attitudes and the young male's attitude toward his father. In
groups where familiarity characterizes the relationship between father and
son, uncle and nephew have a relationship of formal respect; whereas
when the father represents family authority, it is the uncle who is treated
with familiarity.
236
Anthropology: Beyond the Individual Case
Radcliffe-Brown, a keen analyst of kinship structures, had himself
called attention to this phenomenon. But, said Levi-Strauss in response to
Radcliffe-Brown, it does not suffice to study the correlation of attitudes
between father and son and between uncle and sister's son. Rather, playing
structural linguist, one must take into account all the relevant terms and
the relationships among them all. In this particular case, the four crucial
relations are those between brother and sister, husband and wife, father
and son, and mother's brother and sister's son (Levi-Strauss 1963, p. 42).
Even as a linguist studies the phonological relations across many
languages in order to determine the proper set of distinctive features,
Levi-Strauss examined the avunculate in many cultures in an effort to
discover the operative factors. He then went on to propound a structural
law that, in his view, ferrets out the critical factors operating in this
complex set of relationships. The law reads: the relation between maternal
uncle and nephew (whether it be intimate or formal) is to the relation
between brother and sister as the relation between father and son is to that
between husband and wife. According to Levi-Strauss, if one knows one
pair of relations, it is possible, as in any analogy, to figure out the other
(1963, p. 42); he then illustrates his "law" by reviewing supporting exam-
ples. Levi-Strauss goes on to maintain that, even in quite different forms
of descent, one always encounters the same fundamental relationship be-
tween the four pairs of oppositions required to construct the kinship
system. One has thus unlocked the key to such relationships by having
figured out the appropriate unit of analysis. As he says, "This structure is
the most elementary form of kinship that can exist. It is, properly speaking,
the unit of kinship" (p. 46).
Levi-Strauss's early writings on kinship, while certainly controversial,
established his mark as a major anthropological thinker and also promoted
the injection of linguistic techniques (and, to a lesser extent, formal logical
analysis) into the research carried out by anthropologists. Acclaim also
greeted a second wave of work in social organization, another staple inter-
est of anthropologists. Here Levi-Strauss clarified the obscure nature of
dual organizations-where two parallel kinds of clans, often exogamous,
exist within the same village. Levi-Strauss adduced evidence that these
dual organizations actually mask the underlying dynamic force, which
arises from the exchange of women and other commodities. It is this
exchange, rather than the external residence patterns, that reflects the
actual social relations found in the village.
In the early 1950s, Levi-Strauss attended a conference of linguists and
anthropologists where he presented some of the material I have just out-
lined. During his remarks, Levi-Strauss alluded to an "uninvited guest
which has been seated during this conference beside us and which is the
237
I I I THE CoGNITIVE SciENCES : HisTORICAL PERSPECTIVE
human mind" (1963, p. 7). Levi-Strauss felt that in their focus on the material
and social organizational aspects of culture, anthropologists had given
short shrift to the key factor involved in understanding any culture-the
ways in which the human mind takes in, classifies, and interprets informa-
tion. Not coincidentally, Levi-Strauss was pointing out the need for an-
thropologists to consider mental representations just as other pio-
neering cognitivists were beginning to redirect efforts within their own
chosen disciplines-the Zeitgeist was assiduously at work.
Exploring Mind
Levi-Strauss's remark was prophetic concerning his own work, be-
cause for the remainder of his career he has sought to discover the nature
of the human mind in as pristine a form as possible. He has approached
this assignment by studying the ways in which individuals classify objects
and elements, and the ways in which they create and understand myths.
Much of this work is put forth as being empirical-based on the classifica-
tory systems observed around the world and on the myths related in many
Indian tribes. Yet, Levi-Strauss makes no secret of the fact that he must
rely on his own intuitions (which he has dubbed "neolithic"): introducing
his major study of myths he has even declared, "It is in the last resort
immaterial whether in this book the thought processes of the South
American Indians take shape through the medium of my thought, or
whether mine takes place through the medium of theirs" (1969, p. 13).
After all, he believes, it is the same mind-all human minds-in either
case, and the scholar's point of entry is simply not crucial. This may be
cognitive science-but it is a science built on a Cartesian confidence in
one's own mind rather than on the methods of consensual validation
embraced by nearly all other workers today.
While I cannot follow through all the steps of Levi-Strauss's complex
and still uncompleted project, I can say something about the methods he
uses and the conclusions he reaches. In his studies of classification, he
comes down decisively in favor of the proposition that the principal fea-
ture of all minds is to classify, and that primitive individuals classify pretty
much along the same lines, and in the same ways, as the most civilized
persons. He describes the classifying practices of primitive groups as a
science of the concrete: rather than looking for the factors that underlie the
structures or processes of the world (as the trained scientist does), the
primitive mind seeks to classify everyday objects and experiences in terms
of their overt perceptual and sensory properties. These methods do not
always lead to the same categories and classes as those used in the Western
scientific approach-they may be more or less detailed and may have
238
Anthropology: Beyond the Individual Case
differently drawn boundaries; but they reflect roughly the same kinds of
analytic moves on the part of a classifier.
Nor is there an unlimited number of ways in which the human mind
can work. As humans, we are constrained in the kinds of combinations we
can make, in the kinds of distinctive features of opposition with which
we can play. Levi-Strauss laid his cards on the table in this procla-
mation:
The ensemble of a people's customs has always its particular style; they form
into systems. I am convinced that the number of these systems is not unlimited and
that human societies, like individual human beings (at play, in their dreams, or in
moments of delirium), never create absolutely: all they can do is to choose certain
combinations from a repertory of ideas which it should be possible to reconstitute.
(1964, p. 60)
Myth Making
Levi-Strauss's studies of myth making represent his most extensive
search for the rules governing human cognition. In an early work (1963),
he laid out a methodological approach for the structural study of myth.
Proceeding in approved structuralist fashion, he proposed a breakdown of
a myth into component parts or units (the elementary phrases of the myth)
and then the assembling of all units that refer to the same theme or make
the same point. For example, in the case of the Oedipal myths, Levi-
Strauss discerns a set of themes relating to the overvaluing of blood rela-
tions (Cadmos seeks his sister, who has been ravished by Zeus; Oedipus
marries his mother, Jocasta); a set of themes relating to the undervaluing of
blood relations (Oedipus kills his father, Laius; Eteocles kills his brother
Polynices); a set of themes relating to monsters being slain (Cadmos kills the
dragon; Oedipus kills the sphinx); and some unusual names having to do with
difficulties in walking (Labdacos means "lame"; Laius means "left-sided";
Oedipus means "swollen-footed").
Having grouped the various myth themes into these four categories,
"Levi-Strauss then lays out a formula that purports to describe the underly-
ing message of the myth. As he describes it, the Oedipus myth in all its
myriad versions has to do with either the overvaluing or the undervaluing
of the importance of kinship structure, and with the question of men's
239
I I I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTivE
240
Anthropology: Beyond the Individual Case
no longer has to come to terms with objects, it is in a sense reduced to
imitating itself as object" (1969, p. 10). And his examination has confirmed
for him the essential logic inherent in all human thought:
241
I I I THE CoGNITivE SciENcEs : HISTORICAL PERSPECTIVE
cies, his rather cranky set of interests and obsessions, Levi-Strauss will
prove to be an enduring figure in the history of cognitive science. At
mid-century he injected issues of cognition centrally into anthropological
discussions. By invoking the most rigorous approaches of linguistics dur-
ing his time, and applying them imaginatively in the principal domains
within anthropology, he opened up new fields of inquiry. Like Piaget, he
has sometimes aped system, or haphazardly borrowed terms from formal
analysis, rather than being truly systematic: it is therefore up to his succes-
sors to pursue his program in a less personal and more replicable way
(though how to do this is much less clear than, say, in the case of Piaget).
Again, as happened with Piaget, Levi-Strauss's strong positions on an-
thropological issues have served as a convenient point of departure for
revisionist formulations.
When all the criticisms have been considered, few of Levi-Strauss's
particular conclusions may remain; but the critiques will only have been
possible because of his fertile questions and provocative hypotheses.
Moreover, I suspect that one hundred years from now his research program
will be seen as more right-headed than that of his strongest critics-the
true mark of an important thinker. Levi-Strauss will endure because he
posed questions that are central to both anthropology and cognition; out-
lined methods of analysis that might be applicable; and proposed the kinds
of systematic relations which may obtain in such diverse fields as kinship,
social organization, classification, and mythology.
Sperber s Variations
Of those who have some sympathy with Levi-Strauss's enterprise,
Dan Sperber, a former student of his now working in Paris (1968, 1982),
endorses his teacher's once radical program of examining the products of
human mentation and concurs that the model of linguistics is crucial for
anthropologists. Yet, perhaps appropriately for one of a younger genera-
tion, Sperber feels that Levi-Strauss has drawn on the wrong school of
linguistics. The structural approach is as outmoded in anthropology as it
has become in linguistics, maintains Sperber; and it is from the work of
Chomsky, Fodor, and others of the transformationalist school that the
anthropologist must now seek models.
Here the lessons tum out to be largely negative, for language is seen
as a system that works by its own very special rules. Contrary to Levi-
Strauss's vision (and reflecting the shift toward domain-specific princi-
ples), Sperber maintains that linguistic analysis cannot properly be applied
to other cultural phenomena such as myths, custom, and rituals. One
should instead consider cultural phenomena as entities subject to an end-
less amount of mental associations or elaborations, of the sort which go on
242
Anthropology: Beyond the Individual Case
precisely a/fer the usual rule-governed operations of language have been
carried out. It is in the realm of the mysterious, the unanalyzable, the richly
symbolic, that most cultural entities exert their powers-and not in the
relatively lean aspects of language which can be apprehended and analyzed
according to prescribed syntactic, phonological, and lexical regularities.
Sperber points out that most human beliefs are not purely propositional
but are rather semipropositional-not fully logical and far fuzzier. It is
risky to apply to such amorphous belief systems the rigid classificatory grid
of the syntactician or the phonologist. Instead, one needs to study the
processes whereby rich penumbras of meaning are evoked.
Sperber's positive contribution inheres in his characterization of sym-
bolic processes. Rather than being induced or constructed from experience,
the symbolic mechanisms are part of the innate mental equipment which
makes experience possible. These mechanisms of symbolic elaboration,
working in a manner reminiscent of Levi-Strauss's savage, start with the
assumption that the "waste" of the mind ought always to be salvaged
because something can be made of it. It is just because this dross harbors
within it atypical or exceptional conceptual characteristics that it lends
itself to unending symbolic elaboration. In such cases, the symbolic mech-
anism of the mind draws on one's encyclopedic knowledge, one's knowl-
edge of more or less remote categories, and, indeed, any other modes of
information or interpretation that happen to be available-all in an effort
somehow to piece together these disparate elements into an overall sensi-
ble framework. Anthropology is the discipline that has access to the fullest
range of beliefs, practices, and symbolic systems; hence, it is in a privileged
position to lay bare the operation of those human symbolic mechanisms
that supplement the (relatively) pure computational aspects involved in
language, mathematics, and ordinary classification.
243
II I THE CoGNITIVE SciENcEs : HisToRICAL PERSPECTIVE
The way to look at thought is not to assume that there is a parallel thread of
correlated affects or internal experiences that go with it in some regular way. It's
not of course that people don 'f have internal experiences, of course they do; but that
when you ask what is the state of mind of someone, say while he or she is
performing a ritual, it's hard to believe that such experiences are the same for all
people involved .... The thinking, and indeed the feeling in an odd sort of way,
is really going on in public. They are really saying what they're saying, doing what
they're doing, meaning what they're meaning .
. . . Thought is, in great part anyway, a public activity. (Quoted in Miller 1983,
pp. 202-3)
Ethnoscience
What Levi-Strauss undertook pretty much by himself, armed with his own
intuitions, has in the last three decades become in somewhat transmuted
form a major wing of anthropology. I refer here to the field of ethnoscience,
in its various lexical guises-componential analysis, ethnosemantics, cog-
nitive anthropology-all comprising the organized study of the thought
systems of individuals in other cultures and sometimes in our own.
Roots
The factors that gave rise to ethnoscience in the United States in the
middle 1950s resembled those that stimulated Levi-Strauss in France at
about· the same time. Both anthropological circles were affected by the
244
Anthropology: Beyond the Individual Case
example of linguistics as a social science that had achieved unparalleled
rigor. Information theory, stochastic modeling, cybernetics, and computer
simulation were other reminders of the advantages to be gained by formal
approaches to social phenomena. Thanks to the works of the semioticians
like Roman Jakobson (1963} and Charles Morris (1971}, there was increas-
ing awareness of the essentially symbolic property of all human action, and
of the need to explain it in terms of its cognitive, as opposed to its practical,
aspects. According to one of the first ethnoscientific practitioners:
The model for the first experiments in the structural analysis of meaning was
consciously an analogical adaptation of that which had been developed for pho-
nemic analysis. More than one of those who were involved in the early phases of
this experiment have acknowledged that the stimulus to it, as well as the eye-
opener as to how it might be done, came from a combination of their training in
phonemics and their reading of Charles W. Morris' Foundation of the Theory of
Signs. (Lounsbury 1968, p. 223)
245
II I THE CoGNITIVE SciENCEs : HisTORICAL PERSPECTIVE
What he does, instead, is to infer the system of rules that these people are
attempting to apply. He will gain the assurance that he is on the way to an adequate
understanding of these rules from the logical completeness of the system he infers
and from his ability, when using it, to produce behavior that an expert will reward
by saying, in effect, "That's right; that's good; now you've got it." (1968, p. 537)
culminating in the entry after cousin which includes at least twelve possible
relatives. In this particular kinship grid, the terms uncle, aunt, nephew,
cousin, and niece have been used in an extended sense, thus including such
relations as "second cousin once removed" and other distant relations
within an extended family.
The third stage entails a number of observations obtained from the
grid. For instance, all but one of these terms (cousin) specifies the sex of the
relative: some (like grandfather but not cousin) specify generation; all specify
whether the relative is lineally (for example, sons) or nonlineally (for exam-
ple, nephew) related to Ego; and nonlineal terms specify whether all the
246
Anthropology: Beyond the Individual Case
ancestors of the relative are direct ancestors of Ego, whether all the ances-
tors of Ego are ancestors of the relative, or neither.
Now comes the crucial stage: the analyst hypothesizes that three
dimensions will be sufficient to define all the terms. The sex of the rela-
tive (male as a 1, female as a 2 }; generation (b 1 as two generations above Ego;
b2 as one generation above Ego; b3 as Ego's own generation; b4 as one
generation below Ego; b5 as two generations below Ego); finally, linea/-
ify: c1 is lineal; c2 is co-lineal (brother or sister); c3 as ablineal (cousin).
Lineals are persons who are ancestors or descendants of Ego; co-lineals are
nonlineals, all of whose ancestors include, or are included in, all the ances-
tors of Ego; ablineals are relatives who are neither lineals nor co-lineals.
In the next step the terms are now redefined as components. Thus, the
grandfather is a 1 b 1 c1; the grandson is a 1 b5 c1; the sisteris a 2 b3 c2; an uncle
is a 1 b1 c 2 and a 1 b2 c2 ; and a cousin is a b (not marked on either sex or
generation) and c3 . (Note the convention that, when a term does not
discriminate on a dimension, the letter for that dimension is given sans
subscript).
It is now possible to summarize this analysis of kin terms by means
of the paradigm in the following table. Conforming to the technical defini-
tion of a paradigm, each term has been so defined that no term overlaps
or includes another; every component is discriminated by at least one term;
and all terms can be displayed on the same paradigm.
ci Cz CJ
al az aI az aI az
grandfather grandmother
uncle aunt
father mother
son daughter
nephew niece
grandson granddaughter
Source: From A. F. C. Wallace and J. Atkins, "The Meaning of Kinship Terms." Reproduced
by permission of the American Anthropological Association from American Anlhropologisl 62
(1): 58-80, 1960.
247
I I I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
One test of the adequacy of this account, I have said, is that it does not do
violence to my own feel, as informant, for the structure of what is described. This
is the subjective test of adequacy. An equally important test is that it provide an
alien with the knowledge he needs to use my kinship terminology in a way I will
accept as corresponding with the way I use it. This is the objective test of adequacy.
(Quoted in Kuper 1973, p. 573)
248
Anthropology: Beyond the Individual Case
should agree and by which they should ultimately arrive at the same
conclusion. And indeed, at least compared with Levi-Strauss's avowedly
subjective measures, it is possible to train students in componential analy-
sis. In fact, however, componential analysis turns out to be more complex
and "fuzzier" than its originators had hoped.
249
I I I THE CoGNITIVE SciENCEs : HISTORICAL PERSPECTivE
words might have the same meanings. There was the problem of connota-
tion, where words may have the same objective meaning but connote
different affective values (dad, daddy, old man, pop, father, and so on). Other
critics have argued that componential analysis is inherently circular, since
one must begin by assuming the very relationship among terms whose
relationship should actually be fixed only at the conclusion of the investi-
gation. In other words, the existence and coherence of the domain is
presupposed before the investigation begins (Gardin 1965). Some com-
mentators have focused on the enormous problems of translating terms
from a foreign language (particularly one of non Indo-European origins)
into a familiar tongue and assuming that the same kinds of analysis can
be applied to the translations (Crick 1976; Keesing 1976). For example, in
a culture where men and women do few things together, it is risky to
transfer the same meanings to the foreign terms as we would apply to male
and female; and the terms male and female themselves have very different
frequency and usage than do man and woman (Lyons 1968).
How a Navajo thinks and what categories he can employ in his thinking are
not the same.... The assumptions that kin terms have primary ... referents in
all cultures, that kinship and kinsmen are basically the same in all cultures, and
only partitioned differently by various sets of kin terms, and that kin terms corre-
spond to and express the kin categories of a given culture seem extremely naive.
(1971, p. 116)
250
Anthropology: Beyond the Individual Case
with and influence one another. Language as a structured system, not
language as a set of isolated tokens referring to isolated objects, should be
the model of choice for anthropologists.
Perhaps most telling in any discussion about the fate of ethnoscience
is the hegira of Stephen Tyler himself. Initially a staunch proponent of
componential analysis and its first anthologist, he became completely disil-
lusioned in the intervening decade. In 1978 he published The Said and the
Unsaid, in which he spurned the earlier view of language as offering a
reliable window on the cognitive systems of individuals. In his formulation:
No less than the death of meaning should we have forecast from a manner
of thought that emptied thought of all content, and what else could we expect from
a method of analysis that presumed to show that meaning might mysteriously
emerge from the mechanical concatenation of meaningless elements? ... Whether
in art or science nothing is clearer than the intellectual poverty of formalism. (1978,
p. 465)
251
I I I THE CoGNITIVE SciENCEs : HisTORICAL PERSPECTIVE
emotions or diseases, where the line around the domain is not announced
in advance and where an individual's (or a group's) idiosyncratic interpre-
tive system comes more readily to the fore.
In the case of illness, for example, one's responses are so intertwined
with the values of one's culture, the nature of pain, the prevailing theories
of causation and cure of disease, the distinction between observable, and
experienced symptoms that the analyst rapidly loses confidence that di-
verse informants are talking about the same thing or even the same do-
main. Sensing these difficulties, Roger Keesing, another early supporter of
the new ethnography, declared as early as 1976:
For almost fifteen years, cognitive anthropologists have pursued "the new
ethnography" as far as it would lead them. For the last five, at least, it has been
obvious that this would not be very far-that the messianic promises of the early
polemic were not to be realized. "The new ethnographers" have been unable to
move beyond the analysis of artificially simplified and delineated (and usually
trivial) semantic domains and this has discouraged many of the originally faithful.
(1976, p. 307)
It is not by any means the first occasion in cognitive science where formal
methods turn out to work most effectively with those aspects of behavior
that seem least central to mankind's concerns. Clearly, the computational
paradox resonates in anthropological circles.
Stephen Murray (1982), a historian who has reviewed the rise and
subsequent decline of "classical ethnoscience," discerns two different rea-
sons for the evanescence of this form of study. The first stems from a
much-touted promise that was not fulfilled. In the middle 1960s, a major
study of drinking in Chiapas, Mexico, was designed to test the powers of
ethnoscientific research. According to the plan, investigators were sup-
posed to apply the same procedures for analyzing patterns of drinking to
data secured from five villages. However, problems of analyzing the data
were never solved and the investigators themselves eventually moved on
to other problems. As Paul Kay, one member of the team, reports, ·
It turned out that after collecting a huge amount of material and spending two
or three years looking for a set of objective procedures-or even semi-objective
procedures-that could be applied to this material to reduce it to some kind of
logical statement, we gave up, because we could not find such a set of procedures.
Drinking is an institution in Chiapas that permeates the entire lives of a people:
religion, politics, family life, even agriculture is inextricably tied up with drinking,
so to do the ethnography of drinking there is to do the total ethnography. (Quoted
in S. 0. Murray 1982, p. 169)
Kay's colleague Brent Berlin suggests, "We were not convinced that what
could be said from the elicited data was that much more revealing than
252
Anthropology: Beyond the Individual Case
what could be said on the basis of old-fashioned participant observation"
(p. 169).
The experience in Chiapas may have signaled an important lesson for
anthropologists. While certain domains may permit themselves to be
bracketed from the rest of the culture, most domains of interest may be
inextricably bound up with the surrounding context. To study drinking
without studying everything else may (from the anthropological point of
view) be a scientifically untenable posture: in this case, to isolate is to
destroy. Berlin and Kay, who began as enthusiasts of the enthnosemantic
method, are here underlining its limitations. Contextual, historical, and
cultural effects may be of the essence in most of the anthropological
terrain.
Murray cites another reason for the decline of ethnoscience: that is,
this anthropological subspecialty never cohered into a single integrated
perspective but was, at best, a loose confederation. Murray believes that
self-styled revolutionary groups, which demand total commitment and are
organized under a single visionary leader, are more likely to prevail in
academic competition. (The example of Chomsky in linguistics springs to
mind.) A loose confederation offers less resistance to competitors and
ultimately ceases to exist as a recognizable approach. Instead, "having
achieved a measure of recognition and success, ethnoscience grouping
encountered the seeming price of success: splintering.... By the late 1960s
new students were not being attracted and classical ethnoscience was no
more" (1982, p. 172).
There is scarcely any consensus within anthropology about which
steps ought to be taken in lieu of a strict ethnoscience. Some authorities
have continued to carry out studies in the tradition but have substituted
more modest goals. Others have adopted the perspective of Clifford Geertz
(1973, 1983), who feels that anthropologists have erred in attempting to
mimic the natural sciences, and that the anthropologist has greater affinity
with an interpreting literary critic. (I shall consider a Geertzian critique
again in later discussions of cognitive science.) Still others feel that notions
and concepts developed in other areas of cognitive science can be usefully
imported into anthropological study.
Psychological Forays
253
II I THE CoGNITIVE SciENCES : HisTORICAL PERSPECTIVE
254
Anthropology: Beyond the Individual Case
abilities to reason without the usual contextual supports, or to carry
through certain complex chains of reasoning, which seem to develop
chiefly among individuals exposed to years of Western-style secular
schooling.
Investigators have been revisiting some of the issues about language
and thought that had been raised many years ago. While strong experi-
mental support for the Whorf-Sapir hypothesis had never been obtained,
there was a widespread suspicion among anthropologists that the differ-
ences between conceptual systems in remote cultures were vast and that
these might well reflect variations in the structure or the contents of
language. However, in a line of study which continues to exert wide
influence in several cognitive sciences, Eleanor Rosch (1973a, 1973h-then
Eleanor Heider) strongly challenged the Whorfian line: she demonstrated
that, even in cultures with few color terms, individuals still sort, classify,
and otherwise deal with the color spectrum in roughly similar ways. The
language does not affect basic psychological processes. Paralleling this line
of work, Rosch's colleagues at the University of California at Berkeley,
anthropologists Brent Berlin and Paul Kay (1969), showed that the color
terms used by diverse societies all follow a systematic pattern. That is, if
a society has only two color terms, those terms will divide the spectrum
between white and black; if a language contains three terms, then it con-
tains a term for red; if it contains four terms, then it contains a term for
green or yellow (but not both) and so on to the most complete languages
which contain eleven basic color terms (white, black, red, green, yellow,
blue, brown, pink, purple, orange, and gray). While these lines of work
have engendered controversy (as we shall see in chapter 12), they have
helped swing the pendulum of anthropological analysis back to the pole
of universalism: most researchers now believe that individuals the world
over perceive and classify in relatively similar ways, and that the ways in
which they classify reflect the operation of deep principles of mind which
cannot easily be dislocated.
Committed anthropologists thus face a dilemma. On the one hand, as
enemies of racism and cultural chauvinism, they are delighted by the
evidence that individuals the world over appear to think and process
information in similar ways. One of the fundamental puzzles of anthropol-
ogy appears to have been resolved. On the other hand, as scholars who
value the peculiar profiles of different cultures, they do not want these
reassuring signs of universalism to invalidate or render superfluous their
careful study of individual cultures. Thus, they take care to stress that the
identity of mental processes should not in any way lessen the importance
of chronicling and then explaining the vast differences around the globe
in behavior, in patterns of thought, and in the uses to which both are put.
255
II I THE CoGNITIVE SciENcEs : HISTORICAL PERSPECTIVE
256
Anthropology: Beyond the Individual Case
symbolic systems. Every individual must acquire hundreds of thousands
of chunks of cultural information to be competent in the business of
society, and yet the mechanisms whereby these bits are acquired remain
obscure. D'Andrade does venture his opinion that culture is unlikely to be
picked up without considerable modeling and guidance, but that the way
this instruction takes place has not been illuminated thus far by standard
psychological accounts. Here, again, the model of how individuals learn
language presents itself as a tempting, if possibly perilous analogy.
Levy-Bruhl Revisited
In the century since the term culture became formally introduced, and the
discipline of anthropology was first formulated, massive amounts of em-
pirical information have been obtained about individuals in different cul-
tures. We know much more than earlier observers about kin terminology,
social organization, modes of behavior, ways of classification, use of lan-
guage and myth, powers of reasoning. As in many other areas of cognitive
science, the basic scientific dilemma-Do primitives think the way we do?
-has not lost its urgency. But the ways in which scholars think about this
question have become much more sophisticated. Nowadays most inves-
tigators take for granted that the basic modes of perception and classifica-
tion are the same everywhere, but that particular elements in the environ-
ment can affect how-and the extent to which-these processes develop.
Thanks to research in this tradition, many instructive differences among
individuals in diverse societies have been uncovered, even as the funda-
mental continuities in mental processes everywhere seem increasingly to
be confirmed. Here is where future work is likely to proceed.
As anthropologists have tackled the issues vividly raised in Levy-
Bruhl's own writings, they have found themselves engaged in an odd
assortment of activities. There have been flirtations with aggressively cog-
nitive measures-the formal musings of Levi-Strauss and the more pub-
licly verifiable componential analysis of the American school. While nei-
ther has proved a panacea for explicating the thought processes of any
group or individual, let alone an exotic one, clear insights have been gained
about at least some domains in some societies. Thus, Levi-Strauss has
tackled the broadest questions, but has failed to provide sufficiently rigor-
ous descriptions of his analytic methods; in contrast, the componential
analysts have developed relatively precise analytic tools but have not
successfully applied them beyond certain relatively restricted domains.
257
II I THE CoGNITIVE SciENCES : HisTORICAL PERSPECTIVE
258
Anthropology: Beyond the Individual Case
specifics remain all important. Cognitive science can contribute to anthro-
pology, without enveloping it.
While the historical and particularistic aspects of each society con-
tinue to be valued, there is fresh recognition that, in the last analysis, all
that can be attained by any individual in any culture is restricted by the
particular species to which one belongs, and, more specifically, by the
nervous system that one possesses by virtue of one's humanity. For this
reason, anthropologists of every stripe-from Levi-Strauss to Geertz-
have savored discoveries about the human as an organism: the evolution
of the brain, the development of the skeletal musculature system, the
nature of sexual ties at different ages. Such insights from the areas of
biology and neuroscience will not in themselves answer questions about
culture-the levels of analysis are simply too disparate. (Indeed, neuro-
science serves as a kind of "lower bound" to cognitive science and thus is
maximally distant from anthropology.) But, in due course, findings from
the study of the human nervous system may well illuminate how an
individual becomes able in such short order to assimilate and to transmit
to others the practices of the culture in which he or she happens to live.
259
9
Neuroscience:
The Flirtation with
Reductionism
260
Neuroscience: The Flirtation with Reductionism
The Lesion Technique
In carrying out his research, Lashley made extensive use of the tech-
nique of ablation, where specific areas of the nervous system (often regions
of the cerebral cortex) are destroyed by means of a surgical lesion. The
basic goal of the ablation technique is to determine which behavior is
impaired or destroyed following a punctate lesion, and thereby to infer
which functions are typically served by that region of the brain. Lashley
was inclined to be skeptical about the possibility of attributing specific
behavior to specific regions of the brain; but in no way did the apprentice-
ship with Franz foreshadow the extent to which Lashley would go to
pursue this problem or the bold conclusions he was to reach some decades
later. The experiences of the student cannot foretell the achievements of
the mature scientist.
In the decade following his initial training by Franz, Lashley con-
ducted dozens of experiments on the nervous system of the rat. His typical
study involved ablation of an area of the visual cortex of the rat brain and
a determination of the effects wrought on the rat's perceptual powers. By
1929 Lashley was ready to summarize his findings in a major work entitled
Brain Mechanisms and Intelligence. He strongly questioned the significance of
specific neural zones and connections: "It is very doubtful that the same
neurons or synapses are involved even in two similar reactions for the same
stimulus" (1929, p. 3). Describing experiments on maze running after
cortical ablation was examined, Lashley concluded:
The capacity to learn the maze is dependent upon the amount of functional
cortical tissue and not upon its anatomical specialization. . . . The results are
incompatible with theories of learning by changes in synaptic structure, or with
any theories which assume that particular neural integrations are dependent upon
definite anatomical paths specialized for them .... The mechanisms of integration
are to be sought in the dynamic relations among the parts of the nervous system
rather than in details of structural differentiation. (P. 3)
261
II I THE CoGNITIVE SciENcEs : HisTORICAL PERSPECTIVE
262
Neuroscience: 7he Flirtation with Reductionism
acquire the capacity to react in certain definite patterns. Whether the cells
can be mobilized to carry out an impaired function depends upon the
percentage of them still remaining after brain injury, the degree to which
the pattern of behavior has been mastered beforehand, and the strength
of motivation of the animal. Particularly in the case of brain-injured ani-
mals, motivation might have to be quite potent if the organism were to
display its still-preserved abilities.
Lashley's experiments strike a severe blow against any unmodulated
account of the nervous system as a set of discrete centers, each having its
own unique functions, so that destruction of a set of cells would result in
complete loss of a particular function and in total sparing of all remaining
ones. At the same time, he helped to cast doubt on the reflex arc-the bond
whereby each response is triggered by a specific stimulus-which had been
the principal neural model of behavior in higher (as well as lower) organ-
isms. Things were simply not that simple. In study after study, Lashley
deduced the implications of the localizationist position, designed a sup-
posedly critical experiment, and showed that abilities persist despite the
predictions of the localizers, thus giving the lie to those scientists who
claimed that impulses have to be transmitted over certain established paths
in order for specific behavior to be performed.
s
Lashley Iconoclasm
In addition to providing a devastating critique of a simple-minded
localizing position, Lashley also impressed on the next generation of work-
ers the difficulty of coming up with a viable model of the nervous system.
In possibly his best-known paper, the Hixon contribution "The Problem
of Serial Order in Behavior" (Jeffress 1951), which I discussed in chapter
2, he evocatively laid out a whole set of problems that neurobiology had
ignored. Drawing on examples from language, walkin& playing the piano,
and the like, Lashley demonstrated that many sequences of behavior ex-
hibited long planning units which unfold too quickly for them to be
altered or corrected "live." In his view, it was necessary to reconceptualize
current associationist models of the nervous system to allow for effects
that can be manifest for a significant period after initial stimulation. And
so, to choose a single example, in order to understand a double entendre, one
would have to retain in mind (and in the brain) a meaning latent in the
body of the joke, which only becomes "activated" by the punch line. No
simple stimulus-response bonds can explain this behavior: one needs a
model of the nervous system which is hierarchically arranged and features
feedback and feed-forward mechanisms.
Lashley's unorthodoxy manifested itself in the new talk about com-
263
II I THE CoGNITIVE SciENCES : HisTORICAL PERSPECTIVE
puters. At the time of the Hixon Symposium, many noted scholars were
eager to analogize the brain to the digital computer. Lashley voiced a note
of caution:
The brain has been compared to a digital computer because the neuron, like
a switch or valve, either does or does not complete a circuit. But at that point the
similarity ends. The switch in the digital computer is constant in its effect, and its
effect is large in proportion to the total output of the machine. The effect produced
by the neuron varies with its recovery from [the] refractory phase and with its
metabolic state. The number of neurons involved in any action runs into millions
so that the influence of any one is negligible .... Any cell in the system can be
dispensed with .... The brain is an analogical machine, not digital. Analysis of its
integrative activities will probably have to be in statistical terms. (Quoted in Beach
et al. 1960, p. 539)
264
Neuroscience: The Flirtation with Reductionism
assent. For instance, the nervous system turns out to be far more specific,
far less equipotential than Lashley had contended. His belief that the brain
works in a Gestalt-like fashion would find few adherents today. But in
framing the questions sharply, and in introducing the key terms of scien-
tific debate, Lashley for a long time dominated, and still exerts appreciable
influence on, that work in neuroscience that impinges upon cognitive and
behavioral concerns.
The work of Lashley, and of other experimental researchers working
at the intersection of the brain and behavior, raises another issue-and one
crucial for my inquiry here. It has been held by many scientists, especially
neuroscientists, that the optimal way to account for human behavior and
thoughts is in terms of the structure and functioning of the human nervous
system. To some investigators, this neuroanatomical account can comple-
ment accounts proffered in psychological or behavioral language; but for
others, neuroanatomical accounts may eventually render unnecessary ac-
counts in terms of representations, or symbols, or other psychological
argot. In the view of this latter reductionist group, cognitive science
emerges as, at best, a holding operation: a temporary account of mental
activity destined to vanish once an account in terms of synapses can be
attained. And such reductionists find themselves in sharpest opposition to
functionalists-true-blue cognitivists who believe that behavior and
thought must be accounted for completely on the level of representations,
without any regard whatever to the "hardware" in which it happens to be
embodied. The debate about the possibility-and the desirability-of re-
ductionism lurks in the background in any account of neuroscientific work
(see Mehler, Morton, and Jusczyk 1984).
265
II I THE CoGNITIVE SciENCES : HisTORICAL PERSPECTIVE
266
Neuroscience: The Rirtation with Reductionism
involved. While conceding some specificity in the nervous system, Flour-
ens added that "there should, moreover, properly be brought to light
another order of phenomena that includes both this efficacious Unity of the
nervous system that joins all the parts of the system together, in spite of
their diversity of action, and also the degree of influence that each of
these parts contributes to the common activity" (quoted in Hermstein and
Boring 1965, p. 222).
[it) is important for intellectual work. ... [Following ablation] I could perceive
a very decided alteration in the animal's character and behavior .... Instead of as
before being actively interested in their surroundings and curiously prying into all
that came within the field of their observation, they remained apathetic .... They
had lost, to all appearance, the faculty of attentive and intelligent observation.
(Quoted in Diamond 1974, p. 244)
267
I I I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
268
Neurosdence: The RirfaHon with ReducHonism
findings from Gestalt psychology. It was clear that the organism reacted
not just to single stimuli but rather to the relationship among stimuli, to
overall patterns, and to stimuli perceived as part of a given context. More-
over, the Gestalt assumption that the nervous system is organized in terms
of neural fields, operating across wide regions of the cortex, struck a
responsive chord with these neurologists. The opponents of localization in
neurology, like the opponents of atomism in psychology, served as a
bulwark against the elementarist bias of the behaviorists during the early
part of the century. At the same time, holists were far more sympathetic
to the notion that behavior could not be explained satisfactorily in terms
of neural circuitry. As they saw it, there was a continued need for explana-
tion on the psychological level-particularly the global or holistic kinds of
explanation offered by the Gestalt psychologists.
Once again, then, there was a correlation between skepticism about
localization and skepticism about reductionism-not a logically necessary
association, to be sure, but rather a meeting of two ideas in the minds of
many scientists. In the view of these investigators, even if it could be
shown that certain functions regularly break down as a result of damage
to specific areas of the brain, the significance of such a finding was not
clear. Hughlings Jackson, a forerunner of the holist school during the
nineteenth century, had declared that localization of symptoms did not
signify localization of function (Jackson 1932). To give a specific exam-
ple, just because naming breaks down following lesion in the angular
gyrus, it is an unwarranted assumption that naming actually takes place
in this specific site of the brain. Richard L. Gregory has stated this point
crisply:
The removal of any of several resistances in a radio set may cause the emission
of strange sounds, but it cannot therefore be concluded that the function of the
resistances is to inhibit howling .... The southern region of British railways is a
complex system of railway lines, signal boxes, stations and control systems. A
breakdown of a section of the line, a power failure or a slip in the central control
room at Waterloo, may disrupt traffic over a wide area. But we cannot therefore
say that the function of the system is localised in the ... power station, the central
control room ... all are essential. (Quoted in Rose 1973, p. 94)
269
I I I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
270
Neuroscience: The Rirfation with Reductionism
disruption of direct currents and electroencephalographic activities. These
results were difficult to explain on a localizing account; yet, at the same
time, they discredited the major theoretical account put forth by the Ge-
stalt holists. Faced with the results of these experiments, Gestaltist Wolf-
gang Kohler, the chief proponent of field theory, is said to have declared
in desperation "that ruins not only my Direct Current field (theory] but
every other current neurological theory of perception" (quoted in Pribram
1971, pp. 110-11).
271
I I I THE CoGNITIVE SciENCES : HisTORICAL PERSPECTIVE
early in life is flexible, so that it can take place despite the destruction of
large parts of the nervous system; in contrast, later learning depends spe-
cifically on certain developed structures, and there is relatively little plas-
ticity in the system. In this respect, then, it is more plausible to view the
developmental sequence as running from holism to localization. A benefi-
cial effect of Hebb's work was to point up these various complexities and
competing tendencies, making it less plausible for anyone to adopt a rigid
localizing or an inflexible holist position. This intermediary position came
to carry considerable weight and has recently been embraced by computer
scientists who are attempting to simulate vision (see chapter 10).
The need to reconcile these extreme positions was also manifest at the
various pivotal conferences during the late 1940s and early 1950s. During
the Hixon Symposium (Jeffress 1951), the major neurologists, neurophysi-
ologist&, and neuropsychologist& debated with one another the tenability
of the localization position. Psychologist Heinrich Kluver reported aston-
ishingly specific behavior consequent upon lesions in the temporal lobes
of monkeys: these organisms exhibited psychic blindness, strong oral fixa-
tions, an excessive tendency to react to each and every visual stimulus,
profound changes in emotional behavior, a remarkable alteration in dietary
habits, and an increase in the amount and diversity of sexual behavior.
Neurologist J. M. Nielsen described specific forms of agnosia (failure to
recognize objects) which accompany lesions in the occipital lobes. Neuro-
psychologist Ward Halstead discerned predictable mental difficulties in
planning and in abstraction following damage to the frontal lobes.
In response to these strands of evidence in favor of localization, Karl
Lashley reviewed his own long series of experiments on pattern recogni-
tion and maze running in rats. Pointing out that he had found a remarkable
lack of evidence for localization, he taunted Halstead:
272
Neurosdence: The Flirtation with Reductionism
prevents any direct approach to relationally determined facts such as visual
shapes. . . . If we think of cortical function in terms of continuous field
physics rather than of impulses in neurons, the difficulty never arises"
(quoted in Jeffress 1951, p. 65). Such, then, was the situation in the early
1950s: Kohler and Lashley defending a top-down perspective, against
proponents of a view based on specific neural structures. Lines were still
drawn between these two factions, along with a growing feeling that each
of them must hold part of the truth, and a widely discerned need for a more
integrative point of view (like Hebb's) which could give each position its
due.
In the late 1950s, David Hubel and Torsten Wiesel, two young neuro-
physiologists, began to record with microelectrodes from single cells in the
cortex of the cat. Over the next two decades, they were to record impulses
from the nervous systems of numerous cats and other animals as well, and
to do so throughout the visual system at various depths and in other
regions of the brain. For this work, widely recognized as pathbreaking,
Hubel and Wiesel were awarded the Nobel Prize for Medicine or Physiol-
ogy in 1981. Much of our current knowledge of localizing has been ob-
tained from this line of study (Hubel and Wiesel1959, 1962, 1979; Hubel
1979; see also Lettvin et al. 1959).
Hubel, Wiesel, and their associates documented two important
phenomena. First of all, they demonstrated beyond any doubt that specific
cells in the visual cortex respond to specific forms of information in the
environment. The exquisitely organized cerebral cortex contains cortical
columns, and simple cells within the columns respond to such punctate
properties of stimuli as orientation, or presence or absence of light. So-
called complex cells react to lines kept in optimal orientation as they sweep
across a receptive field; perhaps this mode of reaction corresponds to some
early stage in the brain's analysis of visual forms. Some cells respond to
input only from one or the other eye, while others can be influenced
independently by both eyes. For some hypercomplex cells, the most effective
boundary shape is a corner; for others, a tongue shape. Following such
demonstrations, no one could any longer doubt that the nervous system
was highly specific in its mode of functioning.
The second and equally important line of work emanating from the
Hubel-Wiesel laboratory concerns the critical role played by certain early
experiences in the development of the nervous system. While some per-
273
I I I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTIVE
ceptual and motor abilities are clearly "wired in" at birth (or else how could
the organism function well enough to survive?), this is by no m~ans true
for all functions (or else, how could the organism change and learn?). The
visual system of the cat will not develop-in fact, parts of it will atrophy
-if the animal is not exposed to patterned light after birth. Moreover, the
cat must be exposed to a visually varied environment, permitted to use
both eyes, and allowed to move about its environment. If exposed to
horizontal patterns only, the cells normally destined to carry out vertical
processing will either atrophy or will be "taken over" to execute other
functions. It should be noted, too, that the timing of these early experi-
ences can be specific. For example, if, between the third and fifth postnatal
weeks, one of the eyes is prevented from seeing forms, it will become
functionally blind thereafter, even if normal registration and imaging are
restored.
While Hubel and Wiesel's work stands at the center of this line of
research, their efforts have been complemented and strengthened by those
of many other workers. Jerome Lettvin and his colleagues at Massachusetts
Institute of Technology showed early on that receptors in the eye of the
frog are also extremely specific and respond to small round bloblike shapes
which are rather reminiscent of bugs-hence, the term bug detector (Lettvin
et al. 1959). Vernon Mountcastle (1978) documented columns in the
somesthetic sensory cortex that have specific response patterns analogous
to those of the visual cortex. Recording from the inferior temporal lobes
of monkeys, Mortimer Mishkin (1967) finds cells that participate in coding
the physical attributes of visual objects; these cells appear to be involved
in activities much closer to the recognition of objects (and some distance
away from that sensitivity to punctate forms of stimulation which excite
the neurons in the primary visual cortex). There are even cells in the
monkey's cortex that respond maximally to the shape of a monkey's hand
(Gross, Rocha-Miranda, and Bender 1972). For those in sympathy with
specificity and localization of function, the last few decades have yielded
much confirming evidence.
274
Neuroscience: The Rirtation with Reductionism
scientists. Another set of studies carried out at a molar level also bears
upon the claims of the localizers and the holists.
275
II I THE CoGNITIVE SciENCES : HisTORICAL PERSPECTIVE
the two hemispheres themselves differed, with the right hemisphere rela-
tively dominant for more holistic or Gestalt-forms of perception.
276
Neuroscience: The Flirtation with Reductionism
seen in an individual's drawings generally reveals the site of that person's
brain damage. We thus move from a simple assertion that one hemisphere
carries out one function and the other hemisphere a rival function, to the
more sophisticated claim that each hemisphere, or each region within a
given hemisphere, contributes to a given activity in characteristic ways.
Many sentences have been written about the optimal way in which
to describe the dominance patterns of the two hemispheres. Characteriza-
tions range from the mundane (language in the left, spatial function in the
right) to the grandiose (science or rationality in the left, art or intuition in
the right) to a distinction that captures current debates in neuroscience
(localized function in the left hemisphere, synthetic or holistic functioning
in the right). The precise characterization of the functions of the two
hemispheres-or, more probably, determination that such a neat dicho-
tomization is simply not possible--awaits the results of further studies
(Beaumont, Young, and McManus 1984). Yet whatever the mission of each
isolated hemisphere, there is a clearly dynamic interaction between the
two hemispheres. When the left hemisphere is aroused (for instance, by
the sounds of language), it promotes certain kinds of analytic and linguistic
functions. In contrast, arousal or stimulation of the right hemisphere
brings spatial and holistic functions to the fore (see Kinsbourne 1978).
While, in general, it is preferable to sustain an injury to the brain early
rather than late in life, and to exploit the plasticity of that developmental
stage, early is not always better. There are at least three caveats. First of
all, sometimes an early injury manifests no apparent deficits at the time but
produces severe deficits later in life. Thus, injuries to the frontal lobe in
a young monkey may produce no immediate impairment of functioning,
but long-term sequalae become evident at the time when the planning or
mnemonic functions of the frontal lobes would normally mature (Gold-
man-Rakic et al. 1982; Goldman and Galkin 1978). Second of all, when
another area of the brain takes over; the "rescuer" may well sacrifice the
potential for carrying out its own preordained functions. Thus, the right
hemisphere may assume language functions in the left-hemisphere-
injured child, but the child will eventually display spatial deficits, because
of the resulting unavailability of the relevant right-hemisphere zones (B.
T. Woods 1980). Finally, even when another area of the brain assumes a
function, it may not do so in an optimal way. Individuals with only a right
hemisphere do learn to speak, but they end up using different linguistic
strategies, which render them relatively insensitive to crucial syntactic
features of language (Dennis and Whitaker 1976). At least for right-hand-
ers, there may be only one optimal way to learn language-the way that
draws upon intact left-hemisphere structures that have evolved to carry
out phonological and syntactic analyses. Language functions carried out
277
II I THE CoGNITIVE SciENCEs : HisTORICAL PERSPECTIVE
with the right hemisphere depend excessively on semantic and pragmatic
factors; this approach to language proves adequate for most mundane
communication but unsatisfactory for the appreciation of subtle grammat-
ical distinctions.
Even as certain regions of the brain seem preordained to carry out
certain kinds of functions, other considerations militate against a purely
"plastic" perspective. Work in experimental psychology documents that
organisms are "prepared" to master certain behaviors and "counterpre-
pared" to learn other ones. Thus, for example, rats can quickly learn to run
or jump to escape shock but with only the greatest difficulty to press a lever
in order to effect the same escape. Similarly, jumping to avoid a shock
seems a natural or "prepared" response; but if a rat must jump in a box
with a closed lid, learning will be slow and uncertain. It seems reasonable
to contend that normal children are "prepared" to learn language very
quickly, and that at least some children are "prepared" to master the tonal
system of their culture's music with little exposure: it would not be sur-
prising to find children having great difficulty in learning some kind of
"nonnatural" language or musical system contrived by a diabolical experi-
menter or dictated by an omnipotent monarch.
Great scientific debates rarely die out entirely, particularly when there is
a core of validity in the rival perspectives (Holton 1984). Such may be the
case with the controversy between the localizers and holists, which has
been evident since the time of Descartes and highly vocal since the period
of Gall. Studies conducted over the last twenty-five years, in the wake of
the Hixon Symposium, have, however, helped to frame the debate. It is
now conceded that, at least at the level of sensory processing, the nervous
system is specifically constructed to respond to certain kinds of informa-
tion in certain kinds of ways. There is also evidence for "neural commit-
ment" at much more molar levels of representation, even extending to the
two cerebral hemispheres. To this extent, Lashley greatly overstated his
case. On the other hand, impressive evidence continues to accumulate
documenting the resilience and plasticity in the nervous system, particu-
larly during the early phases of development. At such times, even organ-
isms deprived of the usual neuro-anatomical structures are able to adapt
and to carry out requisite functions, sometimes without incurring exces-
sive costs. In this regard, the view of the holists and the mass-action
278
Neuroscience: The Flirtation with Reductionism
proponents remains tenable. As a tentative conclusion, then, it seems that
some localization is accepted by all, but that important islands of plasticity
remain within this general framework.
But while great debates are seldom silenced altogether, they can be-
come muted. Much less energy is devoted nowadays to debates about
localization as against holism (or specificity as against plasticity), at least
in terms of the operation of the nervous system. Instead, studies have
reverted to a more circumscribed terrain. Neuroscientists are devoting the
bulk of their time to the careful study of specific systems in specific
organisms; they are guided by the hope that these systems can be well
understood in their own terms, and that the knowledge obtained thereby
might ultimately inform more general discussions of the neural basis of
cognition, including the controversial issue of reductionism. From the
many examples that could be cited, I shall mention two; these are taken
deliberately from very different levels of analysis and may contribute in
diverse ways to an emerging cognitive science.
279
II I THE CoGNITIVE SciENCEs : HISTORICAL PERSPECTIVE
Thanks to this work, we can now glimpse-perhaps for the first time-
what learning entails at the chemical and neural level. The unsettling
question to be addressed by cognitive science is whether, as other more
complex forms of behavior are similarly described, there will remain a felt
need for a separate explanation at the representational level.
280
Neuroscience: The FlirtaHon with ReducHonism
three months of life, the chaffinch will produce an extremely abnormal
song, which may prove to be little more than a continuous screech. How-
ever, should it be deafened after it has learned its full song, there is no
discernible deterioration in performance.
Bird song is one of the few instances of brain lateralization among
infrahuman animals. Just as the left hemisphere of the brain of humans is
critical for linguistic competence, so the left hypoglossal nerve in the bird
proves crucial for its production of song. One can produce aphasia (or
amusia) in a bird by destroying the left portion of its nervous system. But
the aphasic canary can recover its prior songs because the homologous
pathways of the right hemisphere have the potential of being exploited.
In this recovery of function, songbirds are more fortunate (because their
nervous systems are more plastic) than adult humans.
The work of investigators like Nottebohm and Kandel is based on the
premise that much can be learned at this point through the careful study
of a single system in a single organism. As Peter Marler, another pioneering
researcher on bird song, has remarked: "The research on birdsong learning
is slow and laborious. But I think in the long run this may be the only way
in which we will ever gain proper understanding of the issue that confronts
us ... namely [in this case] the genesis of natural categories in the percep-
tion of animals and man" (1982, p. 93).
To be sure, the two research efforts proceed on somewhat different
assumptions. Kandel hopes that by studying habituation and conditioning,
he will eventually illuminate processes known to occur in a wide range of
organisms, including humans: the assumption here is that certain learning
processes, sometimes modified as "horizontal," cut across all manner of
content (from learning music to mastering drawing), and that these can be
found in relatively analogous fashion across diverse organisms (from ap-
lysia to man). The Nottebohm line of research, on the other hand, proceeds
along a different vein. Bird song is a behavior that clearly exists only in
birds, though it may conceivably have some phylogenetic ties to human
music or to human language. It is, in any case, a self-contained system,
which allows various kinds of experimental manipulation. Interest in bird
song is consonant with a belief that much cognitive activity is "vertically
organized": that there exists a domain called "song" which may well
follow rules different from other domains and is best understood in its own
terms. Any generalizations that may be validly extended from bird song
to other systems in other organisms-or even from one bird species to
another-will only emerge after careful study of these systems on their
own terms. In neuroscience, no less than in psychology and artificial intel-
ligence, a tension can be discerned between a "modular" and a "general
problem-solving" or "central-processing" point of view.
281
II I THE CoGNITivE SciENCES : HISTORICAL PERSPECTIVE
The remarkable success of the work of Hubel and Wiesel and their
collaborators has given fresh impetus to the belief that systems may work
in their own way. In 1978,· David Hubel put it this way:
We are led to expect each region of the central nervous system has its own
special problems that require different solutions. In vision we are concerned with
contours and direction and depth. With the auditory system, on the other hand,
we can anticipate a galaxy of problems relating to temporal interactions of sounds
of different frequencies and it is difficult to imagine the same neural apparatus
dealing with all of these phenomena .... For the major aspects of the brain's
operation no master solution is likely. (Hubel 1978, p. 28)
282
Neuroscience: The Flirtation with Reductionism
view of the brain had been stilled. Not so. Neuroscientist Karl Pribram
argued, in his 1971 book Languages of the Brain, that a belief in specific feature
detectors a Ia Hubel and Wiesel can take one only a limited way: in his
view, the brain is better analogized to a holographic process. Holography
is a system of photography in which a three-dimensional image of an
object can be reproduced (with the appearance of the third dimension
preserved) by means of light-wave patterns recorded on a photographic
plate or film. A hologram is the plate or film with the recorded pattern:
information about any point in the original image is distributed through-
out the hologram, thus making it resistant to damage. Since waves from
all parts of the object are recorded on all parts of the hologram, any part
of the hologram (however small) can be used to reproduce the entire image.
A hologram can store a great deal of information in a small space; indeed
some ten million bits of information have been usefully stored holograph-
ically in a cubic centimeter.
According to Pribram's holographic view, all parts of the brain are
capable of participating in all forms of representation, though admittedly
certain regions play a more important role in some functions, and other
regions are more dominant for other functions. In his view, just as many
holograms can be superimposed upon one another, so can infinite images
be stacked inside our brains. Perhaps when we recall something special, we
use a specific reconstruction beam to zoom in on a particular encoded
memory. Pribram also fixes upon another quality of the hologram: the fact
that it records the same wave front over its surface, repeating it over and
over. Even if only some of a shattered hologram is left, it will still suffice
to reconstruct the entire image (Hooper 1982).
While the holographic analogy has engendered considerable skepti-
cism, many neuroscientists remain sympathetic with Pribram's goal of
showing that the nervous system is not simply a collection of specific
modes of processing; they cling to the possibility that important forms of
knowledge remain, as Lashley had so fervently believed, widely dispersed
throughout the brain. Pribram notes that 11the properties of holograms are
so similar to the elusive properties that Lashley sought in brain tissue to
explain perceptual imaging and engram encoding that the holographic
process must be seriously considered as an explanatory device" (1982, p.
176). Moreover, in a manner reminiscent of the view of limited plasticity,
Pribram himself has recently begun to speak of a 11limited holograph,"
hoping thereby to avoid some of the pitfalls of a full-blown holographic
account. Eric Harth comments:
What interests brain theorists about the hologram is this quality of a distributed
memory (the phrase is Lashley's): Every piece of the hologram says a little bit about
283
II I THE CoGNITivE SciENCEs : HISTORICAL PERSPECTIVE
every part of the scene, but no piece is essential. The other intriguing fact is that
one can superimpose any number of holograms on the same piece of film, and then
reproduce the images of the original scenes one by one without interference from
the others. (1982, p. 88)
The analogy with Lashley's view of the neocortex is striking and may help
explain the continued appeal of a holographic theory of memory.
284
Neuroscience: The Flirtation with Reductionism
system indicates, and as the recent studies of recovery of function in young
animals make clear, the voices in favor of mass action and plasticity have
not been stilled. Moreover, within the specific areas of higher cognitive
functioning (as we shall see), there remain debates about even the most
basic issues. Neuroscientists still do not agree about just how to describe
the functions of the particular cells that Hubel and Wiesel have studied
(for example, in terms of sensitivity to spatial frequency rather than the
detection of line orientation) (Pribram 1980, p. 58); nor how to characterize
the various difficulties that aphasic patients demonstrate; nor how best to
label the functions of the right hemisphere in a normal individual, one who
suffers from unilateral brain injury, or one whose cerebral hemispheres
have been separated for therapeutic reasons. Moreover, even if localization
seems (on the whole) to be more tenable than holism, it is now apparent
that reductionism is a separate issue.
In some ways the neurosciences are different from the other cognitive
disciplines (even as philosophy, the wholly nonexperimental discipline, is
also different). Researchers in the neurosciences stand out from their cog-
nitive-scientific peers because they most closely partake of the model of
the "successful" sciences of physics and biology, because they can most
readily state their questions unambiguously and monitor progress toward
their solutions. While defenders can be found for the propositions that
psychologists or anthropologists have made little progress, or have failed
to define their central issues with sufficient progress, few, if any, informed
observers would level the same charge at the neurosciences. It is for this
reason, in part, that I have avoided an excessively historical approach in
this chapter, and focused instead on one central issue in the field-in order
to show how the question has been framed, to look at the various altera-
tions in perspective, and to evaluate neuroscientific conclusions after a
century of work on the question.
Yet it should be clear that while the neurosciences are different (and
who would decry that difference?), they are not all that different. As I have
observed, fundamental debate continues on many of the central questions.
The uncertainty I have documented reflects the fact that, while rapidly
growing, neuroscience is a young field and is still in the process of d~fining,
rather than resolving, many principal issues. And the uncertainty points
to another consideration as well.
285
II I THE CoGNITIVE SciENCES : HISTORICAL PERSPECTivE
286
Neuroscience: The Flirtation with Reductionism
tion involved in concept formation, but that won't help one bit in under-
standing what a concept is.
From this perspective, it is not possible to enter into the nervous
system as a disinterested observer who is simply chronicling the facts (as
many neuroscientists assume they are doing). Both the topics studied, and
the ways in which they are studied, will reflect implicit theories: theories
about what perception, cognition, or language are; what is important in
each; and how each of these processes occurs. Risks of generalizing from
lower animals to human beings are severe; and, in any event, the less
self-conscious one is about procedures, the more likely it is that one will
make naive errors.
Thus, to take an example from language, a neurologist ignorant of
linguistics might rely on naive intuitions about language: one would there-
fore describe an aphasic patient as unable to use "small words" or to
"speak in full sentences." But a linguistically trained observer will im-
mediately be able to pose questions and introduce distinctions at a subtler
level: Which grammatical categories pose trouble? Are these troubles ap-
parent across different linguistic contexts? Do they correlate with other
phonological, syntactic, lexical, or pragmatic difficulties or distinctions?
Having recourse to such linguistic insights does not guarantee an accurate
analysis. (Sometimes, indeed, too strong an investment in a particular
linguistic theory can blind the observer to counterparadigmatic phe-
nomena.) Yet, to avoid any knowledge of linguistics is equivalent to blind-
ness-by-choice: and so it is now recognized that any aphasia research team
should include psychologists and linguists as well as neuroscientists.
For these reasons, I think it best to regard neuroscience as one of the
border disciplines of cognitive science. Just as anthropology represents a
kind of upper bound for the investigation of cognitive phenomena, so the
neurosciences represent a kind of lower bound. Many of the phenomena
investigated by neuroscientists are either accounted for perfectly ade-
quately without any reference to the representational level or allow but a
subsidiary role for representational aspects: thus, an account of habitua-
tion in Aplysia does not cry out for representational analysis. Such
phenomena do not belong to the mainstream of cognitive science, any
more than do discussions of the religious system of an aboriginal tribe. But
once neuroscientists begin to invade domains that entail more complex
forms of mentation-for example, the domains of language or perception
of objects or logical problem solving-there is no longer any possibility of
finessing these representational issues. It is at this point that interdiscipli-
nary cooperation becomes an imperative. And it is at this point that the
cognitive challenge arises: how best to build explanatory bridges between
the level of the neuron and the level of the rule or the concept.
287
II I THE CoGNITIVE SciENCES : HisTORICAL PERSPECTIVE
288
PART III
TOWARD AN
INTEGRATED
COGNITIVE
SCIENCE: PRESENT
EFFORTS, FUTURE
PROSPECTS
Introduction
In part II, I reviewed highlights from the history of six separate disciplines,
which collectively constitute the cognitive sciences of today and may some-
day be integrated into a single cognitive science. While I have considered
each discipline separately-to monitor how each discipline developed in
terms of its internal standards, the principal issues it confronts, and its
workaday methods-no scholarly discipline operates in a vacuum. There
have been clear conversations among various disciplines-for example,
between philosophy and psychology, between linguistics and anthropol-
ogy, even between artificial intelligence and neuroscience. To put it in the
terms of psychologist E. G. Boring (1950), one can discern the Zeitgeist at
work; or, to use the term of the structuralist historian Michel Foucault
(1970), a common episfeme has been imposed on disparate disciplines.
To convey the factors that seem to have been at work across disparate
disciplines, I shall briefly sketch the history of cognitive science as if it
were a single coherent field. Our prototypical science goes back to the
Greeks-to the writings of Plato and Aristotle, the topics raised in the
Meno. It is here that questions about the nature, the status, the sources, and
the use of knowledge were first raised. The agenda of cognitive science was
fleshed out during the philosophical flowering of the seventeenth and
eighteenth centuries-in the debates between rationalists (like Descartes)
and empiricists (like Locke and Hume) and in the attempts at synthesis put
forth by Kant around 1780 and by Whitehead and Russell around 1910.
Here we see the first intimations of the topics that exercise the cognitive
scientist of our time.
In the nineteenth century, the province of our prototypical science
became that of empirical rather than of introspective scientists. The
unquestioned successes of physics and chemistry, and the inspiring exam-
ple of Darwin in the biological sciences, fanned hopes for a comparable
291
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
292
Introduction
abroad in the period from 1920 to 1950. But as the Hixon Symposium, and
the Macy conferences were being held, as Craik and Turing and Wiener
wrote, and as Bruner, Chomsky, Levi-Strauss, Miller, Newell, and Simon
were pursuing their studies, a cognitivist revolution was brewing. Scholars
were discovering (or rediscovering) the centrality of high-level linguistic
and conceptual activities and the utility of the new tools of computing for
investigating these phenomena. What were once iconoclastic statements
by such scholars became (with surprising rapidity) the new orthodoxy.
Behaviorism was not so much defeated as rendered irrelevant by frontal
approaches to human cognitive processes.
In the late 1950s and early 1960s, there was excitement in the already
established disciplines and euphoria in the new, quintessentially cognitive
discipline of artificial intelligence. But the promise of the cognitive sciences
would not be so readily realized. There ensued serious debates about how
best to approach the discipline: whether by top-down or bottom-up tech-
niques; by general problem solving or expert systems; by programmatic
long-term experimental work, highly selected demonstrations, or detailed
case studies. And when it became clear that early predictions would not
be confirmed, a reaction set in, during the early 1970s: perhaps a cognitive
science was not to be, at least during the lifetime of its most fervent
enthusiasts.
There is another, more attractive possibility, however. Perhaps the
individual cognitive sciences have gone as far as they can within each of
their disciplinary constraints and paradigms. What were once merely polite
and brief conversations among them need to be converted into full-scale
cooperative research efforts on problems central to several of them. Stem-
ming both from a realization that many scientific problems are simply too
complex to be handled by a single discipline, and from a genuine attraction
to the methods and concepts worked out in neighboring disciplines, a
growing number of scientists-many of them raised in the post-1950s
environment-are switching allegiance from single disciplines to the
broader practice of cognitive science and are engaging primarily in inter-
disciplinary pursuits.
Accordingly, in the final chapters of this book, I review several re-
search efforts that qualify for the label cognifive-sdentific. These represent
several disciplines-either combined in one person or laboratory, or in-
volving collaboration across individuals or laboratories-and are intended
to answer the long-standing philosophical questions that originally ener-
gized thinkers in classical times. Of course, many different examples could
have been chosen, and my selection is not meant in any sense to be
definitive. I chose lines of research that involve the central questions in
cognitive science and that are considered by the cognitive-scientific com-
293
II I I TOWARD AN INTEGRATED COGNITIVE SCIENCE
munity to be of high quality. Not beyond criticism (no good scientific work
is immune to critical analysis), this work has rather generated sophisticated
criticism and suggested approaches that may eventually resolve such criti-
cism. Indeed, in what follows, I have not hesitated to cite significant
criticism and, where appropriate, to state my own reservations. I have
chosen to focus on four sets of question that have generated cognitive
science of quality.
In chapter 10, I review contemporary efforts in artificial intelligence,
aided by neuroscience and psychology, to explain how humans perceive
forms and objects. Central to this question is the pathbreaking work of
David Marr and his associates, as well as criticisms put forth by}.}. Gibson
and other believers in "direct perception."
In chapter 11, I consider the status of visual imagery: What is meant
by talk of imagery, and is it proper to treat images as a means whereby
knowledge is represented? My survey here focuses on the work of Stephen
Kosslyn, in psychology and artificial intelligence, and on various philo-
sophical and computational objections to this work put forth by Zenon
Pylyshyn.
In chapter 12, I tum to the issue of how human beings classify objects
and elements in their world. A point of departure are studies of color
naming carried out by the cross-cultural psychologist Eleanor Rosch and
two linguistically oriented anthropologists Brent Berlin and Paul Kay.
Work here impinges on the philosophical issue of whether human classifi-
cations are arbitrary or motivated, as well as on the relationship between
the languages we use and the ways in which we think.
Human rationality is the subject of chapter 13. While philosophers
since Greek times have pondered the nature and the extent of human
rationality, recent work by researchers like Amos Tversky and Philip John-
son-Laird severely questions the model of man as a logical thinker. This
critique, drawn from psychology, linguistics, and artificial intelligence,
has engendered considerable philosophical debate-reason discoursing on
irrationality.
A survey of the best work, and the sharpest critiques of that work,
is one way to evaluate current cognitive science. But, in the end, the way
that the field presents itself-its overall charter-provides an equally im-
portant test. For the most part, I have simply presented this charter and
allowed the field to speak for itself through its history and its work. In the
end, however, I suspend this authorial pose and present my own conclu-
sions about the movement whose early gropings stimulated me to write
this book.
294
10
Perceiving the World
How can we recognize a circle as a circle, whether it is large
or small, near or far; whether, in fact it is in a plane perpen-
dicular to a line from the eye meeting it in the middle and
is seen as a circle, or has some other orientation, and is seen
as an ellipse? How do we see faces and animals and maps
in clouds, or in the blots of a Rorschach test?
-NoRBERT WIENER
295
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
third dimension, but how does this third dimension get recorded on the
retina and then reconstructed within one's head? How different is it to
perceive the world as compared with a picture of the world or with an
image of the world within one's head? Why do optical illusions persist
even after they have been recognized as such? And how do various vi-
sual impressions and images relate to thinking? Are they thoughts in
themselves, do they provide the vehicles of thought, do they reflect the
manipulation of symbolic entities, or are they but epiphenomena, ves-
tiges that do not materially contribute to our ability to know, to learn, or
to understand 7
Questions like these have been pondered ever since philosophy began.
Indeed, the Greeks modeled knowledge upon vision; they invested consid-
erable effort in understanding how we come to know the visible world and
how this knowledge may contribute to--or constitute-general under-
standing. The pre-Socratics were already engaged in debate: Metrodorus
of Chios counseled his followers to disregard evidence from one's senses
and to pay attention to belief; Democritus in turn acknowledged that all
knowledge rests on perception (Barnes 1979). Plato believed that the soul
makes perception possible, while Aristotle was more interested in discov-
ering how the eye actually works. In more recent times, rationalists and
empiricists revisited these issues. Whereas Descartes minimized the impor-
tance of sensory organs, the empiricists saw them as the point of origin for
all knowledge.
Before long, practicing scientists joined these debates. In fact, proba-
bly a majority of psychologists have begun with an interest in perception,
and there has been more unambiguous progress in understanding percep-
tion than in illuminating other mental processes. In my review of psychol-
ogy, I have already touched upon some dominant themes: Hermann von
Helmholtz's belief that there is insufficiently accurate information in the
stimulus itself, and that much of perception accordingly depends upon
unconscious inferences about the scenes that have been observed; J.J.
Gibson's rejoinder that the senses can directly pick up from the environ-
ment information that is needed for survival; the sensationalist's conten-
tion that perception begins with the detection of elementary bits of sensa-
tion, out of which it builds up to more complex objects and forms; the
Gestaltist's contrary assertion that one first perceives overall form, in a
top-down fashion.
Two new lines of research entered discussions about perception. One
line, already reviewed in chapter 9, sought to determine the sensitivities
of particular nerve cells. If the functioning of all neurons and sets of
neurons were to become known, there might be no need for more abstract
or higher-level descriptions of perceptual processing.
296
Perceiving the World
Computer Simulations
297
I I I I TowARD AN INTEGRATED COGNITIVE SciENCE
least handle the whole range of perceptible scenes and the gamut of per-
ceptual tasks-including perception of motion, depth, surface textures,
and other variables that make human perception so powerful. The com-
bination of insights now available from perceptual psychology, neuro-
science, and artificial intelligence made it possible, probably for the first
time, to put forth a reasonably complete account of the early phases of
visual perception-an account that explained how any organism is able to
perceive a world of shapes and objects. The relevant input to this task
came from many quarters, but the pathbreaking conceptions came largely
from David Marr.
298
Perceiving the World
Levels of Scene Analysis
299
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
carry if the sum exceeds 9. Both electrical and mechanical cash registers
follow this type of representation.
In the case of vision, the algorithmic level addresses itself to the
question of the various ways in which a function like stereopsis might
actually be represented and carried out by some mechanism. In this partie~
ular case, Marr and his colleagues had laid out an elegant set of procedures
for computing stereoscopic vision; it turned out, however, that this set bore
no resemblance to the processes apparently used by the brain. Marr
wanted his algorithms to be consistent with what is known about animal
perception: accordingly, that algorithm was replaced by one more consist-
ent with evidence from psychophysics and from neuropsychology.
The third level concerns the device in which the process is to be
realized physically. The same algorithm can be implemented in quite dif-
ferent technologies. Addition can be run on various electrical or mechani-
cal machines but can also be carried out in the brain.
Any task may be achieved by various algorithms, and any algorithm
is susceptible to many realizations in a given hardware. The decision about
which way the algorithm will actually be realized is made by each investi-
gator, and many students of artificial intelligence are unconcerned about
the ways in which human beings happen to accomplish visual perception.
Marr' s interest in the level of realization clearly concerned the possibility
of building computer programs to parse scenes effectively. This was the
essential"existence" proof that an algorithm actually worked. As we have
seen, however, he let his work be guided by the procedures that seemed
to be used by the human brain-possibly because these seemed most likely
to work effectively for inorganic machines as well.
Marr noted that it was easier for scientists to work on the algorithmic
and the mechanical realization level-levels that lend themselves to actual
experimentation. But in his view, the level of computational theory was
most important to tackle at the present time. Here Marr reflects his belief
that the nature of the computations that underlie perception depends more
on the computational problems that have to be solved by any system, than
on the particular hardware in which the solutions happen to be imple-
mented in the most familiar instances.
Marr cautioned against excessive concentration on the nature of brain
processes alone; as he noted, even total knowledge of anatomy and physi-
ology would not allow one to understand why neurons have receptive
fields. To understand how the neurons of the visual system actually ac-
complish their tasks, one must draw upon mathematical principles in-
volved in interpreting images. Yet Marr was conscious that no discipline
in itself could unravel the mysteries of perception. A true cognitive scien-
tist, he once declared:
300
Perceiving the World
The moral is that ignorance in any of these three fields can be damaging. Just
as the modern physicist has to know some mathematics, so must the modern
psychologist, but the psychologist must also be familiar with computation and
have a clear idea of its abilities, its limitations, the fruitful ways in which to think
about processes and, most importantly, what it takes to understand these processes.
(1982, p. 187)
301
I I I I TowARD AN INTEGRATED CoGNITIVE SciENcE
occur in a scene at the point where edges and changes in surface contour
are likely to occur. This phase ends with a representation that makes
explicit the size and disposition of intensity changes-allowing one to
detect the boundaries in an image and what has caused them. The primal
sketch consists of a set of blobs oriented in various directions; these are
reminiscent of the sorts of features discerned by Hubel and Wiesel's detec-
tors-contrasts, spatial extent, general orientation at a local level. All of
these reductions and simplifications are conceived of as mental representa-
tions or symbolic depictions of the "raw information" transmitted by the
light: perception consists of a series of such simplified sketches en route to
a more veridical view of the world.
302
a. b. c.
,..
.........,
..\.:. 'J
.. .
. "
•" '
An image (a), the spatial components of its primal sketch (b), and a
reconstruction of the image from the primal sketch (c) .
•
• ··-·········· ·········· •
• -f·.·.:···t ···-~·-····~·······t r··:::r . . ...
·~.-"II t '-....;-. •
·~_ .. 1"11
'
', .......... ,_ •
• ··........... ~ ..... l!....... l ......, .... ·"····o:;.;·····.
• •
rn r~~--~ -f----L_o_w_e_r_A_r....,m
~
Arm
~- Hand
~-1---------1
~
I
'--------i-------'
tified from an image which capture what is specific about the objects in
question.
What does this mean in practice? The object is broken down into
components and subcomponents until all of its parts have been uniquely
specified. The model's coordinate system and component axes must be
identified from an image, and the arrangement of the component axes in
that coordinate system need to be specified. Whereas the products of a
primal sketch look like line segments oriented in various directions, the
products of the three-0 sketch look like stick figures composed of pipe
cleaners. This reason is that according to Marr and his colleague Keith
Nishahara (1978), the brain automatically transposes the contours it has
derived from the ZY2-0 sketch onto axes of symmetry that resemble stick
figures. By the time the three-0 sketch has been constructed, the final
result should be a unique description of any object one can distinguish; the
same object should always yield the same unique description no matter
what the angle of viewing; and different representations should reflect
the similarity between different objects, while also preserving whatever
differences may matter.
We have, then, a series of steps through which presumably both
humans and machines must pass in making sense of a scene or an image.
The first computational stage, the formation of the primal sketch, consists
of description of a scene in terms of a vast collection of features like edges,
lines, and blobs-the kinds of feature that may depend upon specific
neural detectors a la Hubel and Wiesel. This primal sketch, an initial
symbolic representation of the image, is formed by processing mechanisms
that are completely independent of any "high level" knowledge about
objects. The second stage involves analysis of the primal sketch by sym-
bolic processes that are capable of grouping lines, points, and blobs to-
gether in various ways. The point here is that generally one can see a round
triangular bulb before knowing that it is a chestnut tree. And, correlatively,
in certain varieties of brain damage, one may be able to see shapes quite
reliably without knowing what objects they represent. Then, in the final
stages, an actual identification of an object, along with its component parts,
is made; and this identification should uniquely determine which object is
being perceived. Top-down knowledge about the nature and construction
of the objects of the world is presumably brought to bear in this last phase
of early visual processing. And so, according to this scheme, the sorts of
knowledge about the world which had earlier appeared essential for per-
ception actually come into play only after shapes have been completely
analyzed.
I have sought to convey the boldness of Marr's approach in artificial
intelligence and, without going into technical details, to suggest the way
304
Perceiving the World
in which he conceived the steps of visual processing. It is equally important
to convey the enormity of the enterprise. At the level of the primal sketch,
for example, the primitives the algorithms must detect include blobs, ter-
minations and discontinuities, edge segments, virtual lines, boundaries,
groups, curvilinear organization, and other such elements. Marr himself
attempted to describe the process in general terms:
One initially selects roughly similar elements from [the image] and groups
and clusters them together, forming lines, curves, larger blobs, groups, and small
patches to the extent allowed by the inherent structure of the image. By doing this
again and again, one builds up tokens or primitives at each scale that capture the
spatial structure at that scale. Thus if the image was a close-up view of a cat, the
raw primal sketch might yield descriptions mostly at the scale of the eat's hairs.
At the next level the markings on its coat may appear-which may also be detected
directly by intensity changes-and at a yet higher level there is the parallel-stripe
structure of these markings .... at each step the primitives used are qualitatively
similar symbols-edges, bars, blobs, and terminations or discontinuities-but they
refer to increasingly abstract properties of the image. (1982, p. 91)
305
II I I TowARD AN INTEGRATED CoGNITIVE SciENCE
306
Perceiving the World
If neurophysiology was a theoretical vacuum when he entered it, it is now
seething with lively controversy about the validity of his ideas on the visual
system .... Even if no single one of Marr's detailed hypotheses ultimately survives,
which is unlikely, the questions he raises can no longer be ignored and the method-
ology he proposes seems to be the only one that has any hope of illuminating the
bewildering circuitry of the central nervous system. David Marr's lifework will
have been vindicated when neuroscientists cannot understand how it was ever
possible to doubt the validity of his theoretical maxims. (1982, p. 992)
Reactions to Marr
It is too soon to evaluate which portions of Marr's program are likely
to survive, and which will need to be modified or altogether scuttled. Most
commentators have expressed admiration of Marr's research program and
awe at the amount he was able to accomplish with his colleagues in a few
years, but have generally withheld public comment on which particular
aspects of his effort are least robust. Conversations with scholars know-
ledgeable about vision have revealed areas of likely criticism.
Like many other pioneers, Marr expressed his convictions in a strong
and clear-cut manner, even as he did not hesitate to criticize others. In
reaction, not surprisingly, observers stress the ties between Marr's contri-
butions and those of other workers (for example, Hom 1975) as well as the
307
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
In Gibson's view (1950, 1966, 1979; see also Turvey et al. 1981), organ-
isms are so constituted, and live in a world so constituted, that they will
readily gain the information they need to survive and to thrive. In partic-
ular, our sense organs are designed to pick up information from the ex-
ternal world. Thus, when one detects the third dimension, the relevant
spatial information is simply presented in the light, without one's having
to infer distances or to correlate information from eye and hand; there is
no need for the kinds of unconscious inference which scientists from
Helmholtz on have proposed. Initially in life the information will be
308
Perceiving the World
relatively gross, but with time and experience, it will be increasingly
fine-grained; still, in any. case, the information will be adequate for sur-
vival, veridical, devoid of misleading cues. The information is available
in the world and needs only to be picked up. There is no need to operate
upon it or process it; there is no need to draw on prior knowledge, on
mental models, on interpretive schemata (Gibson 1967). If researchers
understood how people attend to what is there, all the problems of per-
ceptual psychology-indeed, perhaps all the problems of psychology
proper-could be satisfactorily explicated.
On which evidence did Gibson draw in coming up with this radically
simple orientation to perception? In the course of many years of ex-
perimentation, he became extremely impressed with the kinds of informa-
tion available to the visual sense alone. For example, according to classical
empiricist theory, the sense of touch was deemed central to an organism's
ability to perceive depth. But Gibson showed that there is sufficient infor-
mation in the visual sphere to allow the discrimination of depth. The
gradient of texture density is an important clue. Consider a piece of paper
or a linoleum floor with a recurrent checkerboard pattern: one sees the
density of the texture as invariant across the entire array. So long as one
peers straight ahead (or straight down) at the pattern, the density of the
texture does not change. But as the stimulus slants, or as one peers at it
from an angle, the density of texture changes from the near edge to the
far edge, and the viewer receives precise and unambiguous information
about the distance of each part of the stimulus from one's eyes (Hochberg
1978). Hence attention to texture gradient suffices to yield information
about the sizes of objects on a surface or the arrangement of surfaces with
respect to each other: there is no need for the use of other sensory modali-
ties, the calculation of ratios, or other unconscious inferences.
Gibson also stressed the important contribution to perception made
by a person's motions in the world. So long as one is forced to sit passively,
any scene will appear ambiguous. But if you are free to walk about,
changes in the optic array will be precisely tied to the voluntary move-
ments of your body. As you continue to explore, information is routinely
obtained and that information in tum yields more relevant information.
Moreover, the changes in the optic array that result from motions initiated
by the individual make it very easy to figure out what is occurring in the
visual world. Thus, active exploring individuals exploit a perceptual
system that is maximally informative about space and distance.
Based on these and many other observations and considerations, Gib-
son arrived at extreme skepticism about the whole computational ap-
proach. He objected to the notion of mental representations, mental opera-
tions, the processing (as opposed to the direct "pickup") of information, and
309
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
310
Perceiving the World
son once declared, "that invariance comes from reality, not the other way
around. Invariance in the ambient optic array over time is not constructed
or deduced; it is there to be discovered" (quoted in Royce and Rozeboom
1972, p. 239).
It is here that cognitive scientists of almost every stripe have locked
horns with Gibson. As Marr phrased it, the detection of invariants is
exactly and precisely an information-processing problem, which must be
approached by the tools of modem psychology. Gibson greatly underrated
the difficulty of such detection. The only way to understand how the
detection works is to treat it as an information-processing problem.
311
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
312
Perceiving the World
beliefs, goals, purposes, and other intentional states directed toward their
percepts. As Fodor and Pylyshyn point out, it is not good enough that we
directly perceive that a rock can be used as a weapon (because it is "grabba-
ble11 and "hurlable11); we need an account of how apprehension of such a
property can occur without inferences. There is no "neutral11 information
available to a perceiver; either the information is in the world but not
available in a relevant form to the perceiver; or the information is inter-
preted by the perceiver and hence can no longer be deemed neutral
(Fodor 1984).
Thus, when Gibson talks about the light conveying information con-
cerning a layout, he is actually talking about semantic relations-the infor-
mation is about something; but he has no way of indicating how one
recognizes such semantic relations. The question how one gets from prop-
erties of light to the properties of the layout has only one conceivable
answer in the Fodor-Pylyshyn view-through inferential mediation. The
pressing question for the cognitivist is to understand the sorts of empirical
consideration relevant to deciding which properties of the light are attended
to, interpreted, and then acted upon.
In Fodor and Pylyshyn's view, Gibson's longtime focus on problems
of visual perception led him to underestimate the difficulty of constructing
a cognitive psychology that dispenses with mental representation. The
prototypical perceptual relations are extensional: they have to do with the
information about features of the environment, such as those in an image.
Here the need for internal representations and inferences may be less
patent. But most prototypical cognitive relations-like believing, expect-
ing, thinking, and so on-are intentional, and the constructs of mental
representation are required to explain intentionality. There is a big differ-
ence between seeing x and seeing x as y, and what is central to cognitive
psychology is the ability to see something as an entity-a rock as a tool
or a weapon or a stool rather than just a blob.
Fodor and Pylyshyn (1981) conclude that one needs either an inde-
pendent account of the meaning of a representation, as is called for by the
Establishment view, or of the specification of a property, as required by
Gibson. The problem of accounting for the meaning of a representation
may be tractable, since the meaning of a representation can perhaps be
reconstructed by reference to its function; but Gibson gives no indication
at all of how to specify a property, no explanation of how a configuration
of light can ever specify such interpreted properties. And so "where the
Establishment line offers, anyhow, a pious hope, the Gibsonian line offers
only a dead end" (Fodor and Pylyshyn 1981, p. 192). Missing the point
about inference, about mental representations, about intentionality are all
thus aspects of missing the same point.
313
II I I TowARD AN INTEGRATED CoGNITIVE SciENCE
An Aggressive Defense
With such a staunch attack from the M.I.T. trio of Ullman, Fodor, and
Pylyshyn, * one might wonder whether the followers of Gibson would
retreat (or retrench) in silence. Far from it. In their commentaries on Ull-
man's target article in The Behavioral and Brain Sciences, and in a response to
Fodor and Pylyshyn even longer than the original article, Michael Turvey
and Robert Shaw, two Gibsonian psychologists at the University of Con-
necticut, indicate that the cognitivists have missed the point of their own
position. To start with, Turvey and Shaw remind us of the extent to which
Gibson's work has illuminated fundamental questions of perception and
has helped to explain the perception and the behavior of a wide range of
organisms moving about in a wide range of environments. The proof is in
the pudding, they say, and they reject out-of-hand the notion that the
Gibson approach is as vacuous as Skinner's: they then wrap the mantle of
Chomsky around themselves by claiming, "Just as Chomsky used the
regularity and ease of natural language acquisition as a fact to justify
treating language as a special subject matter, so Gibson and his followers
have argued for the importance of doing justice to natural, effective
perception" (Turvey et al. 1981, p. 239).
According to Turvey, Shaw, and their assoc_iates (abbreviated here-
after as Turvey and Shaw), Gibson's ecological approach seeks to account
for an organism's apprehension of its environment and the manner in
which it controls its actions with respect to that environment. The ap-
proach focuses squarely on organism-environmental relations, never on
what might be in the organism's head. Indeed, many issues can simply be
bypassed: for Gibson, "awareness" is always "awareness of some prop-
erty," and so there is no need to posit inferences or symbolic calculations.
Secondary become the issues of what counts as perception, of whether
perception should be construed as judgment, of whether perception is
direct or indirect, and of how inference figures in the scheme of things.
Once again turning Fodor and Pylyshyn's argument against them, the
Turvey team concludes that the Establishment's views are insufficiently
constrained and have placed a burden on inference which it simply cannot
bear. Instead, Turvey, Shaw, and colleagues argue for a conception of
natural law that posits meaningful relations between organisms and envi-
ronment. By "natural law," they are referring to scientific principles-in
this case, laws that explain why organisms perceive and behave in the
ways they do by virtue of their fit with their environment. The use of the
term perception should then be restricted to relations captured by such laws.
*Pylyshyn, currently at the University of Western Ontario, is a frequent visitor and
sometime collaborator at M.I.T.
314
Perceiving the World
Ecological theorists like Gibson seek to extend the application of
natural law as far as possible, because that strategy promises tight con-
straints and an explanation of lawful evolution of inference in principled
ways. In contrast, Establishmentarians like Fodor and Pylyshyn want to
extend cognition and intelligence as far as possible and thereby to limit the
role of natural law. Regularities, then, must be accounted for by mental
rules and representation-which are themselves constrained by very little
except unsystematic intuitions (Turvey et al. 1981, p. 245). The Establish-
ment talks a great deal about how to make the right inferences and very
little about how to locate the right premises. The ecological school provides
a way to get the right premises by laying out the laws of perception; it
derives these laws of nature by careful observation of, and experimenta-
tion on, organisms as they pick up information from and operate effec-
tively within their natural environments. The Establishment has to gamble
that it can guess how an organism is going to know what to do in a given
situation; the ecological school has the much more straightforward task of
simply finding out what sorts of things this kind of organism actually
does in this kind of a situation.
The role of natural law centers on affordances and effectivities. Gib-
son's "ecological science" studies the relations between affordances-things
that are grabbable, climbable, and the like-and effeclivities-things that can
do the grabbing, climbing, and so on. The examples used by the Gibson
school account for how organisms carry out these and countless other
activities in their natural environments in order to survive and thrive.
According to this scientific credo, psychologists are called upon to describe
the lawful regularities of the environment if they wish to produce a scien-
tific explanation of the origin, function, and causation of behavior (Turvey
et al. 1981, p. 274). This tack includes laws about everyday objects and
events, which have the affordances that govern behavior of importance to
the organisms of the world. And so, the "ecological psychologists" fill their
articles with analysis of places where wasps can lay their eggs, elements
that sharks can eat, stems on which marsh periwinkles can climb, and the
like-no optical illusions or ambiguous sentences for them.
Turvey and Shaw deplore the rampant tendency to ascribe to an
organism neural detectors or structural descriptions that essentially repre-
sent the very property that is sensed. This practice permeates the Estab-
lishment: for Z to see, detect, register a property x of X, Z must "have"
property x in some sense, neurophysiologically or conceptually. In contrast
(as Turvey and Shaw have put it), the ecological school eschews rules (or
computation) in favor of natural laws, representations in favor of occurrent
properties, and concepts in favor of affordances. Instead of trying to stuff
properties into the heads of organisms, the ecological school keeps these
315
III I TowARD AN INTEGRATED CoGNITIVE SciENCE
properties where they belong-in the natural world. All told, the ecologists
make central to their science the discovery of natural laws governing the
organism's relation to its environment, rather than cognitive laws, which
carry out operations on mental representations.
Contrasting Perspectives
As the ecologists describe it, we are left with two sharply contrasting
images. The Establishment treatment of the intentionality problem con-
jures up an image of an organism (such as the hermit crab), on the occasion
of becoming hungry, moving about with a concept of food in mind and
looking for something that will match the concept; or an organism (such
as the marsh periwinkle), on the occasion of impending danger from the
approaching tide, moving about with a concept of a thing that can be
climbed up in mind and looking for something in the environment that will
match up with that concept. The ecological approach conjures a very
different image-of an organism, on a given occasion, moving in the con-
text of one set of nested laws rather than another. Confronted with the fact
that a gannet can dive accurately into the water and scoop up a fish when
hungry, the ecological school asks about the various kinds of laws that can
control the gannet's behavior and how the values for the gannet's dive are
arrived at in accordance with some law when it dives for food. The latter
image expresses belief in a natural basis to intentionality, whereas the
Establishment image does not (Turvey et al. 1981, p. 299).
In the view of Turvey and Shaw, there is at stake a larger philosophy
of science issue. Are the uniformities observed in nature reflections of an
underlying concrete framework of laws, or are they only the insidious
invention of the human mind? Turvey and Shaw criticize Kant's arrogance
in saying, "The understanding does not draw its laws from nature but
prescribes them to nature" (1981, p. 299), and prefer to draw constraints
from biology and physics rather than from the more elusive realm of the
human mind. These researchers prefer to believe that there is a natural
order, which it is the scientist's role to discover-just as it is the organism's
role to discover the natural properties in its environment that will give it
all that it ever needs to know.
The Gibsonians attack the Establishment in terms of its conception of
scientific truth. While both Gibson and his critics are involved in a scien-
tific effort designed to figure out how perception and cognition occur, the
Gibsonians are likely to find the crucial information in the environment
316
Perceiving the World
and the organism's relationship to it; the Establishment cognitivists pay
attention instead to the presuppositions and biases built into the organism
and to the way information in the world becomes transformed or recon-
structed upon its apprehension by the organisms.
It is possible to take the position that these are simply two views of
the same basic situation: one paying far more attention to the environment;
the other paying far more attention to the organism; but both attempting
to explain the same set of phenomena in parallel ways. One can also divide
the territory to be explained, leaving simple perception to the ecologists,
complex inferencing to the Establishment. Both parties occasionally take
such positions, when they are in a mediating mood; but in general, the gulf
seems much deeper. The very terms that are central to the Establish-
ment-in/en/ions, inferences, schemas-are rejected by the Gibsonians as un-
necessary: even as the terminology favored by the latter--alfordances, elfecHoi-
fies, in/ormation pickup-is deemed by the Establishment to be vacuous or
insufficient. The fundamental scientific themata are also different: Gibson
reflects a belief in the real world as it is, with all the information there, and
the organism simply attuned to it; the Establishment reflects a belief in the
constructive powers of the mind, with the external world simply a trigger
for activities and operations that are largely built into the organism. De-
pending upon whose view holds sway, cognitive studies will look quite
different.
From one vantage point, the dispute between the Establishment and
the ecological school can be depressing. Here we are, two thousand years
after the first discussions about perception, several hundred years after the
philosophical debates between the empiricists and the rationalists first
raged, and leading scientists are still disagreeing about fundamentals.
Though the current debate cannot be mapped directly onto other debates
-nominalist versus realist, empiricist versus rationalist, unconscious in-
ference versus sensory registration versus "pickup" of relevant informa-
tion-the themes are familiar enough, and the arguments frequent enough,
as to make one question whether there has been progress.
Yet once one transcends these statements of position-in very spare
form-one finds far more agreement. Both schools certainly believe that
our knowledge of the mechanisms of perception has been enhanced over
the past decades. We know much more now about how perception of the
world comes about than we did fifty years ago, and a good deal of credit
for this knowledge accrues to Gibson himself. Disagreements about a
research program lie in whether the study of organisms and environments
can suffice or whether an additional layer of analysis is necessary. Marr and
his colleagues have made a strong case for the proposition that perception
can only be understood if the nature of the problem involved in vision is
317
I I I I TOWARD AN INTEGRATED COGNITIVE SCIENCE
Possible Reconciliations
318
Perceiving the World
mind a new wave of artificial-intelligence attempts to study the process of
visual perception. Inspired by Marr's example, but diverging from it in
significant respects, a sizable number of scientists are now investigating
visual-processing systems modeled closely on the primate nervous system
(Ballard, Hinton, and Sejnowski 1983; Brown 1984; Feldman 1981; Gross-
berg 1980; Hinton and Anderson 1981; Hinton, Sejnowski, and Ackley
1984; Hofstadter 1983; Rumelhart and McClelland 1982; Waldrop 1984).
These approaches-variously called "parallel visual computation," "nee-
associationism," "nee-connectionism," or "massive parallel processing
systems" (M.P.P.S.)-begin with a critique of standard computer simula-
tion of vision, which uses serial"von Neumann-style" symbolic process-
ing. In their view, it makes far more sense to simulate vision by using
(physical or virtual) machines involving many independent processors
(perhaps as many as a million) carrying out many processes at one moment
in time. Such parallel systems take the Marr scheme much further than he
explicitly went. To the notion of different modules carrying out their own
separate analyses, one now adds the notion of many units operating and
exchanging information in ways analogous to many brain cells or columns
firing simultaneously. Computation is performed by excitatory and inhibi-
tory interactions among a network of relatively simple neuronlike units,
which compete and cooperate so that certain units become active and
others are suppressed. Eventually, thanks to statistical properties of the
ensemble, the network settles into a state that reflects its particular "task"
-for example, perceiving a given image.
In these M.P.P.S.'s, memory and perception occur in a distributed
fashion: that is, instead of there being a single central control, or the
complex passing of information between modules, many units operate
simultaneously and achieve their effects statistically. The multiple connec-
tions allow much of the knowledge of the entire system to be applied in
any instance of recognition or problem solving. There are other advantages
to a distributed representation. No information inheres in a specific locus;
thus, even though many units (or cells) may be destroyed, the relevant
memory or concept continues to exist. Also, because of the widespread
distribution of information, it is possible to arrive at a decision even if a
match turns out to be noisy, or incomplete, or approximate. These proper-
ties seem closer to the kinds of search and decision organisms must carry
out in a complex and often chaotic natural world.
An important feature of the M.P.P.S. approach is that it dispenses
with some of the staples of artificial intelligence as the latter has been
customarily conceived (and as it most annoyed Gibson). Central to the
classical Newell-Simon view is the positing of symbolic structures, upon
which operations are performed in a specified order, as the result of a
decision procedure. In this new dispensation, it may be possible to dispose
319
I I I I TOWARD AN INTEGRATED COGNITIVE SCIENCE
320
Perceiving the World
entailing two different systems: one, massively parallel, engaged in such
probablistic endeavors as object recognition: the other, sequential, dedi-
cated to symbolic manipulation, rendering dichotomous judgments, and
engaged in such deterministic activities as logical problem solving (Kosslyn
1984; Fahlman, Hinton, and Sejnowski 1983).
How does the new approach of M.P.P.S. speak to the issues being
debated by the Gibsonians and the Establishmentarians? Paradoxically,
the parallelists are using the mechanisms most clearly associated with the
Establishment view-powerful electronic computers-in order to put forth
a view of perception closer to that embraced by Gibson. He did not live
long enough to pass judgment on this new approach to the simulation of
vision. But several features of the new approach-its fidelity to the me-
chanics of the brain, its spuming of complex symbol manipulation or
intricate decision procedures about what step to carry out next, its impor-
tation of vast amounts of real world knowledge, and its suggestions about
how Gestalt phenomena might emerge from the competition and coopera-
tion of various neural networks-have a Gibsonian ring to them. More-
over, to the extent that such efforts actually result in an effective simula-
tion of vision, in a way that seems concordant with the human nervous
system, the need for theoretical discussions about the "right way" to think
of perception may become increasingly academic. (But see Fodor 1984 for
an alternative view.)
In any case, if either of these rival schools-or one like them-turns
out to be successful, then the problems of visual perception will have been
largely resolved. The Marr-Ullman approach will represent more of a
triumph for cognitive science: it is a self-constructed dialogue among sev-
eral participant disciplines, no one of which can hold the answer to the
puzzles alone; and it draws explicitly on cognitive science concepts such
as representation and symbolic operations. If the Gibsonian approach were
to triumph, it would call into question the need for much cognitive-science
baggage; it would constitute more of an explanation in a classical psy-
chobiological or psychophysical framework. The M.P.P.S. approach falls
somewhere in between: the technology of cognitive science in the service
of a view of perception which is closer to neurology than to psychology.
In my own view, Gibson's work is a logical place in which to begin
the study of perception: his ecological perspective provides vital informa-
tion concerning the phenomena of perception and the kinds of information
in the environment to which any perceptual apparatus must be sensitive.
In Marr's terms, Gibson helps us to understand the nature of the computa-
tional task involved in perception. But Gibson's work proves of limited
help in understanding the actual steps involved in perceiving specific ob-
jects. In contrast, Marr has put forth a plausible account of how an orga-
321
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
nism may actually proceed from intensities of light to the parsing of the
objects in a scene. He has provided a mechanistic account of perception
which seems internally coherent and consistent as well with converging
evidence drawn from several cognitive disciplines. And even if his particu-
lar account turns out to be wrong or incomplete, or if the approach of the
parallel processors carries the day, he has defined the likely parameters for
future debates about early visual perception. So, in the terms I have used
in this chapter: even if Marris wrong about the extent to which bottom-up
analysis can occur before top-down factors have to be invoked, he has
made a persuasive case for the importance of the symbolic or representa-
tional level of information processing. As I see it, the burden of proof is
now on the Gibsonians.
Even if Marr (or the parallel-processing school) is right, and we can
account for the perception of objects with relatively little recourse to
knowledge about the external world, still it is patent that the recognition
of objects, or even groups of objects, marks just the beginning of the
cognitive enterprise. Any student of cognition must be concerned with
what happens to these initial perceptions when they are invoked in plan-
ning, problem solving, or simple recollection: one must concern oneself
with the way in which organisms determine the identities of objects, how
to classify them, and how to reason about them. These even more chal-
lenging puzzles will occupy the following chapters.
322
11
Mental Imagery:
A Figment of
the Imagination?
Now we have already discussed imagination in the treatise
On the Soul and we concluded there that thought is impossi-
ble without an image.
-ARISTOTLE
323
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
324
Mental Imagery: A Figment of the Imagination?
Jacqueline Metzler (1971) exposed subjects to geometric figures and asked
them to indicate as rapidly as possible whether the two figures were
actually representations of the same object as seen from different vantage
points.
But, strikingly, the difficulty of the task (as measured by the time it
took to give a response) proved readily predictable in terms of the number
of degrees by which the second figure had been rotated. Thus, a figure that
had been rotated eighty degrees took longer to specify as identical to the
target than one rotated fifty degrees, but a model rotated one hundred
degrees proved even more difficult to specify as identical to the target than
had the former two. It was tempting to conclude that the subjects were
mentally rotating these figures, and that the greater the distance of the
rotations, the longer the road to a correct response. Moreover, there was
no need just to conjecture, because the subjects themselves verified this
account. It seemed reasonable--indeed, almost inescapable--to conclude
that human beings generate mental images of these forms and rotate them
through some as yet undefined mental space.
According to Stephen Kosslyn {1980), possibly the major contempo-
rary student of imagery, Shepard's findings caused a sensation in the
cognitive community. The data were striking and systematic, apparently
reflecting a basic capacity of the cognitive system. Moreover, these elusive
internal images turned out to yield a psychophysical law that was simple
yet robust: time to judge identity (or non-identity) was a monotonic func-
tion of the physical distance between the forms. One of the least tangible
constructs in psychology had yielded a scientific law of striking precision.
There were other reasons for the excitement. In light of Shepard's
results, it made sense to think of an individual as having pictures in his
head. He never employed this expression, which cannot be construed in
a sensible manner anyway; but he made respectable the idea of an analogue
mode of mental representation, a mode capturing certain of the relations
of proximity that can also be perceived in the physical world. Shepard's
results called into question current efforts to explain all of thought in terms
of one kind of computational mechanism-that of the serial, digital com-
puter which processes one kind of information. The typical approach of
this era had held that information is represented in the brain in lists or
networks of propositional information. But this approach-that what we
know consists of lists or of propositions-seems utterly inadequate to
account for the mental rotation findings and phenomena. Instead, it made
more sense to think of the brain as passing through an ordered series of
states, which mimic ordinary processes when physical stimuli are being
observed in the world. Imagery should be thought of in its own terms,
rather than as a cryptic version of verbal mediation or symbolic manipula-
325
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
tion. Perhaps there were two separate and equally valid forms of mental
representation. And perhaps a study of the less familiar imagistic mode
might help to clarify some of the later stages of visual perception-and do
so in a way different from either a neurological or an ecological perspective.
Taking off from these and other equally beguiling findings, Kosslyn and
his colleagues have undertaken three major and coordinated lines of work
(Kosslyn et al. 1979). First of all, they have added to the file of empirical
evidence that favors the existence and flexibility of mental imagery. In a
set of extremely ingenious experiments, the Kosslyn team has delineated
major properties of the imagery system.
To convey the Kosslyn approach, let me introduce a representative
study (Kosslyn, Ball, and Reiser 1978). Subjects are shown a map contain-
ing seven fictional locations-a rock, a tree, a beach, a patch of grass, a well,
a hut, and a lake. After having had a chance to familiarize themselves with
the map, one is asked to imagine the map in one's mind and then to answer
various questions about it. For example, one is asked to focus on one
location in the map and then to look for a second one. Assuming that the
second location is on the map, a subject is directed to picture a little black
speck moving as rapidly as possible from the first to the second point and
then to push a button. When one cannot find the named second location,
one is asked to push a second button.
Kosslyn found that the time to scan from location A to location B was
a linear function of the distance between the two sites on the map. Appar-
ently the subjects were scanning a mental image, just as one would scan
the physically present map. As a check that the subjects were not simply
using a propositional representation or a mental list of the locations, sub-
jects were seen in a second condition. Now one was simply asked to
indicate whether a named location was on the map. It turned out that,
under this condition, there was no hint of a distance effect. Evidently
subjects just consulted a list in this instance. Thus, it seems most unlikely
that subjects were employing some propositional representation in the first
study: for, if they had been, the results on both experiments should have
been identical (Kosslyn 1983, p. 46).
What of other results demonstrated by the Kosslyn group? It takes
more time to survey an image of a large object than an image of a small
one; more time to survey an image that travels in a third dimension than
326
Menta/Imagery: A Figment of the lmaginaHon?
one that only cuts across a flat plane; more time to see small details on an
image than large ones; and it is more difficult to imagine objects known
to be hidden or concealed by a barrier than ones known to be in a visible
position. Ever sensitive to the possibility that subjects may be reasoning
(rather than imagining) their ways to the correct answer, Kosslyn and his
colleagues have paid particular attention to experimental conditions where
subjects could not have known the correct answer. They have discovered
that subjects perceive imagistic effects that even scientists could not pre-
dict from a verbal description of the task, but that individuals placed in
the actual concrete situation will reliably experience. On the other hand,
there are some effects-for example, certain afterimages-that individuals
will perceive when confronted with actual displays but that turn out not
to be experienced when one is asked to imagine the customary eliciting
circumstances. These latter lines of research support the independent va-
lidity of imagery. If individuals always behaved as their knowledge dic-
tated-using images just in those cases where they thought that they should-
then they would not display such paradoxical effects.
Having demonstrated a massive number of results putatively due to
some form of mental imagery, Kosslyn and his colleagues have gone on to
devise a comprehensive theory of this capacity. While not boldly claiming
that human beings have pictures in their heads, these researchers defend
the notion of a "quasi-pictorial" form of mental representation called
"imagery." In their view, this form of mental representation is as important
for an understanding of cognition as is the more usually invoked proposi-
tional form.
Here the plot thickens. Many psychologists would be content to allow
imagery back into the psychologist's lexicon and even permit it to be used
as a "local explanatory construct" for certain reliable findings. But when
talk shifts to imagery as a basic property of human cognition-as a primary
way in which information can be symbolized or represented-then psy-
chologists and other cognitivists grow much more cautious. After all, if as
elusive a concept as an image is allowed to serve as a psychological expla-
nation, then on what basis can other "pretenders" be excluded? For their
part, philosophers also begin to get nervous: What does it mean "to have
an image" or to read information off an image? What can an image in the
head be anyway? (Block 198la).
According to the Kosslyn account, images have two major compo-
nents: the surface representation is the quasi-pictorial entity in the active
memory that is accompanied by the subjective (Kosslyn's term) experience
of having an image. The images are likened to displays produced on a
cathode-ray tube by a computer program operating on stored data. In other
words, the images are temporary spatial displays in active memory that
327
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
Computer SimulaHon
Nowadays, in psychology, when one has a detailed description of a
process, it is desirable (as well as fashionable) to attempt to develop a
computer model of it-as Kosslyn and his collaborators have set out to do.
They posit two kinds of data structures. The first consists of a surface
matrix which represents the image itself. This image-data structure is repre-
sented by a configuration of points on a matrix. The matrix corresponds
to a visual buffer, a spatial medium used to support the representations that
make possible seeing during perception as well as during imaging. A quasi-
pictorial image is achieved as a person selectively fills in certain cells of this
matrix. The image depicts information about spatial extent, brightness,
and contrast; it has limited resolution, causing contours to be obscured if
an object is too small. The matrix corresponds to short-term visual mem-
ory; and, therefore, representations within it are transient and can be
maintained only through effort.
The second set of data structures consists of the long-term memory
files, which represent the information used in generating images. One kind
of representation stores visual information about the literal appearance of
an object, such as how something looks; included is information on the
object as viewed in a number of different ways. The second type of long-
328
Mental Imagery: A Figment of the Imagination?
term representation is a set of facts about the imaged objects. These facts
are represented discursively, in a propositional format: included is such
information as the names of the most highly associated superordinate
categories and a classification of the object's size. These sets of data struc-
tures are then used in three imaging processes: routines for generating
surface images, for classifying these images or parts thereof, and for trans-
forming images.
In a highly schematic example, Kosslyn and his colleagues illustrate
how these data sources and processes get mobilized in a program. For
example, in generating a detailed image of a chair, the IMAGE procedure first
constructs a skeletal image of the chair and then searches for factual
information stored in long-term memory for the names of parts that go in
the chair. It may find, for example, that chairs have cushions, and that
these are the foundation part of the chair. PUT and FIND procedures then
locate the relevant part of the image (the seat) by means of a set of
procedures describing the seat. Once the seat has been found, a FIND opera-
tion passes back Cartesian coordinate information of where this part fits
the image. Once the location and size are computed, an appropriate PICTURE
routine is called, and the part is integrated into an overall image (Kosslyn
et al. 1979, p. 542).
A chief reason for actually programming the computer, as opposed to
merely producing flow charts, is that it helps one to discover the conse-
quences of some claim and to study interactions among components (Koss-
lyn 1980). The fact that the simulation works also solves the homunculus
problem: the program performs the integration, and there is no need for
a "little man" to read the images. In some cases, Kosslyn's simulation has
actually brought to light problems that were not anticipated, and has
accordingly inspired a revision of the model. Thus, for example, in the
original simulation there was no provision for scanning to the inactivated
region of the surface matrix. Only when the program was asked to find an
object in a certain image, and received an error message, did Kosslyn make
needed changes in the program. In addition to calling attention to deficits
in the model, the computer simulation has also suggested various new
studies; these have led to further changes in the model which could be
incorporated into further revision of the simulation as well.
In many ways, the Kosslyn program can be considered a prototypical
cognitive-scientific enterprise. First of all, it addresses a long-standing set
of philosophical issues via a systematic program of experimental research.
It deals unabashedly with the level of mental representation, while avoid-
ing many of the weaknesses of earlier attempts to posit and utilize con-
structs of mental imagery. Rooted in the first instance in psychology, it
relies heavily on a model of computer simulation. In addition, as we shall
329
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
see, there are also significant ties to work in neuroscience and in philoso-
phy, and these are being vigorously explored at present. Many aspects of
image generation-including its similarities and differences from ordinary
perception or ordinary memory-have been clarified. It is not that the
Kosslyn group has resolved all outstanding issues about imagery: in fact,
much of the work and many of the claims remain controversial. Yet,
building on the work of Shepard, Paivio, and other researchers, Kosslyn
has made the study of imagery a respectable topic within cognitive science
and illuminated crucial facets of this mode of mental representation.
Imagery is now central in any cognitive map of the discipline.
330
Mental Imagery: A Figment of the Imagination?
the effects they reported: in this case, imagistic explanations need not be
invoked. Accordingly, Ronald Finke, another associate of Kosslyn, studied
a phenomenon-the perception of complementary colors as afterimages-
about which nonpsychologists are ignorant (Finke and Schmidt 1978).
Despite their ignorance, subjects asked to imagine the phenomenon re-
ported precisely the same afterimages as do na"ive individuals confronted
with a "real life" presentation. Finally, Kosslyn's overall position has been
bolstered by Martha Farah's finding (1984) that at least one of the compo-
nents of his model-the image-generation component-can be destroyed
in isolation by damage to the brain. Moreover, and in surprising contradic-
tion to what the literature on brain laterality suggests, the image-genera-
tion component seems to be found in the posterior portions of the left (and
not the right) hemisphere of the human brain.
There have been other criticisms as well. Some scholars have ques-
tioned the meaning of key terms in Kosslyn's model and have voiced the
charge of theoretical looseness. For example, the term quasi-pictorial, as
applied to an image, seems a clear case of wanting to exploit the connota-
tion of picture without paying the price for it. There have also been general
criticisms about the ecological validity of Kosslyn's research program. Ulric
Neisser expresses this point of view:
Why does the theory suggested here strike the reader as clever rather than
insightful, as cute model making rather than serious psychology. I think it is
because the thinking of Kosslyn and his collaborators is completely detached from
everything we know about human action or perception.... It attempts to "account
for" a sharply restricted body of experimental results (usually reaction latencies)
by relating it to an equally restricted class of models (usually computer programs
of something similar). (Quoted in Kosslyn et al. 1979, p. 560)
331
I I I I TowARD AN INTEGRA TED CoGNITIVE SciENCE
332
Mental Imagery: A Figment of the Imagination?
In explicating intrinsic properties, Pylyshyn has coined the termfimc-
tional architecture to refer to basic information-processing mechanisms of the
system for which a nonrepresentational account suffices. The functional
architecture includes the basic operations permitted by the biological sys-
tem, as well as such built-in constraints as the size of the memory, the
capacity of a buffer, the list of permissible operations, and the like. The
functional architecture cannot vary in ways that demand a cognitive expla-
nation (in terms of goals, beliefs, representations, and the like). Rather,
differences in cognitive phenomena can themselves be explained by ap-
pealing to the different arrangements that can obtain among the fixed set
of operations.
The notion of functional architecture allows Pylyshyn to cleave apart
two fundamentally different types of processes: those whose explanations
require appeal to certain kinds of representations, and those whose expla-
nations do not. Those processes requiring appeal to the representational
level are termed cognitively penetrable: they can be (and routinely are) affected
by an individual's symbolic processes, including beliefs, wishes, and the
like. Countenancing inferences, they fall outside the functional architec-
ture. In contrast are those capacities that are cognitively impenetrable: part of
the functional architecture, these processes are carried out in an automatic
and encapsulated way, impervious to an individual's particular beliefs.
Impenetrable capacities can be likened to the hardware built into a com-
puter; penetrable capacities are programmable, hence subject to change.
It would be desirable if the human cognitive repertoire were largely
impenetrable: then one could explain many processes as occurring neces-
sarily, courtesy of neural wiring. It turns out, however, that much of
human cognition is penetrable. And, in particular, in Pylyshyn's view,
imagery turns out to be cognitively penetrable: rather than being a requirement
of a certain kind of cognitive system (a part of the hardware), it can be
changed in all kinds of ways simply by one's giving (propositional) in-
structions to oneself. If the results reported by Kosslyn were supposed to
contribute to our knowledge of how the mind works, these functions must
be cognitively impenetrable: they must always apply when imagery is
putatively involved. But one can, in fact, claims Pylyshyn, change such
functions: they are cognitively penetrable. Put concretely, you can make
your attention jump from place to place in an image as easily as you can
make your scanning speed change--evidence that these activities are at the
behest of one's ideas and beliefs.
Pylyshyn revisits Kosslyn's map-location task but interprets it in a
different vein. Where Kosslyn believes that a subject is focusing on one
point after another in a mental image, Pylyshyn believes that the subject
is simply imagining the "real life" situation and how one would behave
333
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
The notion of a discrete atomic symbol is the basis of all formal understand-
ing. Indeed, it is the basis of all systems of thought, expression or calculation for
which a notation is available .... No one has succeeded in defining any other type
of atom from which formal understanding can be derived. Small wonder, then, that
334
Mental Imagery: A Figment of the Imagination?
many of us are reluctant to dispense with this foundation in cognitive psychology
under frequent exhortations to accept symbols with such varied intrinsic properties
as continuous or analogue properties. (1984, p. 51)
335
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
A Wittgensfeinian Critique
Pushing this line of criticism yet further is William Shebar (1979), a
philosophy student who worked with Kosslyn and then subjected some of
his ideas to a Wittgensteinian critique. As I noted in chapter 4, Wittgen-
stein had little regard for psychological theorizing: in his view, concepts
were often adopted unwittingly from certain seductive features of our
language. Thus, a Wittgensteinian approach calls for careful attention to
the language involved in such expressions as "imagining" or "having an
image," and is skeptical about images as an explanatory construct within
psychology.
As Shebar puts it, a careful look at the language involved in talking
about images reveals the many traps concealed therein. One should not
336
Mental Imagery: A Figment of the Imagination?
think of an image as a thing, for what kind of a thing is it? It is preferable
instead to talk about the action of imagining, which is something that
individuals legitimately do. You can tell someone to imagine a house-
because that is something that people can do-but you cannot tell someone
to "see a house," for seeing is a different kind of action from imagining,
and not one subject to voluntary control. By the same token, it is equally
misleading to analogize an image to a picture. One needs materials to make
a picture but not to make an image; yet this form of locution leads us to
think of imagery as a private slide show which we can see on the private
screens of our minds. Thus, you obviously cannot draw someone else's
picture, but you do have the capacity to imagine it.
Shebar objects to Kosslyn's characterizations of his results. The same
results can be explained without the need for an analogy with imagery.
Anyone knows that it takes longer to scan further distances on real pictures
and thus, when asked to imagine this process, conceives of it along the
same lines. The fact that it takes a longer time to press the button does not
imply the existence of a particular process called imagery-or a particular
element called an image-which is longer; it is simply an affirmation that
whatever happens when one scans a longer distance takes more time.
In Shebar's view, the term infernal representation is loaded from the start
with connotations of external representations on which it is modeled. But
it is not proper to talk about a hidden picture as a representation-for to
whom would it represent anything if one could see it? Shebar senses a
pernicious circle. Subjects are understanding the instructions given by
Kosslyn by grasping the very analogy that has been used to make the
predictions and is later used as an explanation for the result. As Shebar
puts it, a subject's understanding of the "inner process" talk in the instruc-
tions depends upon one's grasping of the "outer" process talk on which
it is modeled; and thus, one is bound to produce just those regularities that
look like evidence for the theory.
Shebar's critique is potentially devastating for imagery studies and
possibly as well for other areas of psychology which dabble in internal
representation. From his perspective, psychologists think they are study-
ing processes but are actually examining the effects of adapting a certain
way of talking about things. The researchers are using a form of locution
that actually reflects pretheoretical ways of discussing experience, rather
than mechanisms that can actually be investigated and understood. Ac-
cording to this account, the problems of psychology will be solved not by
newer data or more precise terminology but, rather, by the realization that
they are not genuine problems.
In my view Shebar's critique is tonic, perhaps slightly toxic, but not
as fatal as he might wish. There are, in fact, considerable risks in adapting
337
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
338
Mental Imagery: A Figment of the Imagination?
may well be that nonlinguistic animals have images; and there is little
doubt that they perceive in ways similar to human beings.) But any cogni-
tive science worth its name will eventually have to offer accounts as well
of those cognitive capacities that are heavily infiltrated or penetrated by
linguistic and conceptual processes. In entering this area, cognitive scien-
tists confront the most vexing issues within their science-and also those
issues about which human beings care most deeply.
339
12
A World Categorized
In defining the human being as a featherless biped, the Greeks were not
simply providing a succinct definition of their own species. They were
exemplifying their belief that common objects of the world can be clas-
sified into groups, and that these groups can be defined by certain criteria!
attributes. Thus, all human beings belong to a single class called humans,
and that class can be defined reliably by the whimsical phrase to which I
have alluded.
Consider the set of accomplishments inhering in this terse formula-
tion. While all living creatures respond in similar ways to entities they
deem similar, these responses are typically unreflective. Noting that
humans everywhere group entities together constitutes an important
achievement. Realization that these categorizations can form a nested hier-
archy is a further insight. For instance, a person treats different views of
a terrier as instances of the same terrier; groups together all terriers; then
groups terriers with poodles into the categories of dogs, dogs and cats into
the category of domestic animals, and on up the hierarchy to mammals,
vertebrates, animals, living things, and, simply, entities (Keil1979). More-
over, in addition to forming and organizing categories, human beings also
ferret out the traits defining each category. If one can supply a satisfactory
definition of a terrier, a dog, a domestic animal, and the like, one will then
have a rationale for placing every entity into one or more appropriate
categories.
From the days of Aristotle, such practices of naming, defining, and
categorizing have undergone philosophical scrutiny. By the middle of this
century, a certain position had become entrenched as the "right way" to
think about categories, concepts, and classifications (a trio of terms I shall
use here interchangeably). And yet in the past thirty-five years, during the
very period when cognitive science has been in the ascendancy, this view
340
A World Categorized
of how we categorize the world has undergone the most severe attack, until
today virtually no one holds it in its pure form.
In this chapter, I shall first review the major points of the classical view
of concepts and then consider the major attack on this position in the fields
of psychology, anthropology, and philosophy. Today it is no exaggeration
to say that the classical view of concepts has been replaced by a natural view
of concepts. The new view is not without its critics, including a fair num-
ber of commentators who yearn for the less ambiguous days of the classical
view, as well as radical scholars who doubt altogether the utility of invent-
ing categorization. After surveying a range of responses to the natural
view, I shall take stock of what has now been established about human
classificatory practices, and how this knowledge can inform an emerging
cognitive science.
341
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
The processor was assumed to be rational, and attention was directed to the
logical nature of problem-solving strategies. The "mature western mind" was
presumed to be one that, in abstracting knowledge from the idiosyncracies of
particular everyday experience, employed Aristologian laws of logic. When applied
to categories, this meant that to know a category was to have an abstracted clear-
cut, necessary, and sufficient criteria for category membership. If other thought
processes, such as imagery, ostensive definition, reasoning by analogy to particular
instances, or the use of metaphors were considered at all, they were usually rele-
gated to lesser beings such as women, children, primitive people, or even to nonhu-
mans. (Rosch and Lloyd 1978, p. 2)
In one sense, this classical view of categorization was consonant with the
aura of cognitive science-a science built upon the unambiguity of the
computer. But just as the general view of the computer as a single all-
purpose logical machine eventually gave way to a plurality of computa-
tional stances, so, too, this general view of classification was not able to
withstand the force of arguments and empirical data from several quarters.
A strong blow against it came from work in a domain the classical view
had initially adopted as its own-the area of color naming.
342
A World Categorized
naming and labeling colors, and these means of subdividing the spectrum
seem quite diverse. Some cultures have just two or three names for describ-
ing all colors; while other societies, including our own, feature a family of
terms-unmodified (blue) as well as highly modified (light sky blue). Since
each culture apparently carves up the color sphere as it sees fit, there are
apparently no natural laws at work here. Rather, one culture decides to cut
the color spectrum at a certain point (for example, at wavelength r ); while
another culture divides it at wavelength y, or divides it a different number
of times. The task of a person within a culture is to learn the color name
that that culture has arbitrarily hit upon, just as one learns the name of
family members, flora and fauna, and various human artifacts, ranging
from tools to forms of government.
One scholar who had been thinking along these lines was Roger
Brown, psycholinguist at Harvard University. Working in the early 1950s
with his graduate student Eric Lenneberg, Brown had become interested
in the way in which a culture's particular parsing of the color spectrum
may affect how individuals from that culture classify and later recall spe-
cific hues. In a review written in 1975, Brown has caught the Zeitgeist of that
period: he has also suggested the ways in which it came to be undermined
by Eleanor Rosch, who happens to have been his student nearly twenty
years later.
As Brown recalls in his review, he and Eric Lenneberg (1954) wanted
to test the Whorf-Sapir hypothesis: that is, one's conceptualization of the
world reflects the particular terms and concepts of one's culture (see also
chapter 8, p. 235). These researchers selected color for two principal rea-
sons: first, because colors can be described in an objective "culture-free"
way, drawing upon an extensive psychophysical apparatus; second, be-
cause the cross-cultural records documented many cultures with disparate
color lexicons.
For various reasons, the original study by Brown and Lenneberg actu-
ally involved only native speakers of English. Subjects were shown some
twenty-four colors and asked to name them: those colors that were readily
named were called "codable." Another group of subjects was then briefly
exposed to a small set of colors and thereafter shown a large set of hues
and asked to indicate which they had seen before and which they had not.
The colors in the new set included ones that had been rated as codable and
ones that had not been so labeled. The results provided support for a weak
version of the Whorf-Sapir hypothesis: specifically, subjects were some-
what better able to recognize those chips that had been rated as highly
codable than those that had not. The authors judged that the existence of
names within the American lexicon had exerted an influence on the sub-
jects' behavior, by making those colors that were readily named easier to
recognize.
343
I I I I ToWARD AN INTEGRATED CoGNITIVE SciENCE
344
A World Categorized
study of the effects of language on thought, the color space would seem
to be a prime example of the influence of underlying perceptual cognitive
factors on the formation and reference of linguistic categories" (Heider
1972, p. 20).
Why, then, was this set of findings damaging to the classical view of
concept formation? For many years, it had been assumed that the lines
between colors were arbitrary, drawn as seen fit by a culture; individuals
would simply mirror these boundaries in their own classificatory and mne-
monic behavior. Now Rosch was calling into question these various lines
of argument. The naming practices turned out to be incidental. The ways
in which individuals from different cultures remember colors seemed to
reflect the organization of the nervous system, not the structure of particu-
lar lexicons. Certain colors are "good" instances of a color because of the
physiology of the human visual system and not because of specific naming
practices. Indeed, a lexicon codes aspects of color that are already salient
rather than making these aspects salient.
Rosch was by no means content simply to conduct studies of color
naming among an exotic population (Mervis and Rosch 1981; Rosch 1977,
1978). Following her return to America, she probed a wide gamut of
domains and concluded that the story on color is highly pertinent else-
where as well. According to the classical view of classification, based
primarily on the use of artificial stimuli, each category is defined by a finite
list of criteria! features: members display these features; nonmembers do
not. But this classical picture does not apply to the world of natural objects
-like birds-nor does it prove particularly illuminating in the realm of
numerous man-made objects, like furniture or tools. In the world of every-
day reality, highly correlated (non-independent) features prove the rule.
For example, given the capacity to perceive feathers, furs, and wings, a
perceiver soon comes to realize that in the empirical world, wings occur
with feathers more than they do with furs. There is, in other words,
considerable redundancy in the appearance of members of the same cate-
gory-not the independence of features posited in the classical view. Rec-
ognition mechanisms exploit these redundancies.
There are other intrusions from the world as it actually is. For exam-
ple, with respect to man-made entities, objects with one sort of appearance
are more likely to be grouped together as chairs (because they possess the
potential for being sat upon) while another group of objects is more likely
to be grouped as drinking vessels (because they possess the potential for
being held and poured from). While some characteristics may appear to
demark chairs and others to demark drinking vessels, one looks in vain for
a set of defining criteria; instead, the kinds of action these objects seem to
elicit (or afford) constitute a more useful aid to classification.
345
II I I TOWARD AN INTEGRATED COGNITIVE SCIENCE
346
A World Categorized
but harbor prototypes, with less prototypical members being apprehended
with reference to the extent that they resemble (or fail to resemble) the
prototype. Categories exhibit internal structure and that structure yields
psychological consequences: basic-level objects tum out to be the most
readily named and recalled; less psychologically accessible, but still useful
and necessary for various purposes, are the finer-grained level of the sub-
ordinate and the overarching level of the superordinate. Finally, thecate-
gories do not have firm boundaries: many members sit astride the border
of two or more categories.
Several factors may have contributed to this state of affairs in the
world of categorization. Having evolved over many years to be able to deal
efficiently with their environment, human beings tend to group together
into categories entities that appear similar to their perceptual apparatus or
call upon similar actions, or both. Moreover, what looks similar is not in
the least arbitrary. For example, our visual system is so designed as to treat
certain reds as being better than others, and to draw the line between red
and orange, or between orange and yellow at specific points on what looks
(to the instruments of the physical scientist) like a continuous spectrum.
This view makes clear contact with what is known about the physiology
of color vision, and links are forged as well with recent work in logic: the
new variety of fuzzy set logic, which deals with degree of membership in a
class, turns out 1:o fit well with the claims being made by Rosch and her
colleagues. And so one telltale finding about color naming in a remote
Stone Age tribe has fostered a rev9lution in the way we conceptualize
concepts.
While nearly ~veryone! agrees that the Rosch findings undermine a
strong view of t~e classical theory of concepts, critics question whether the
natural view, taken alon~, can replace it. In one study, Sharon Armstrong,
Ula Gleitman, and Henry Gleitman (1983) examined the structure of a
category that was clearly defined in the classical criterial way: the category
of odd number. To their delight, these authors determined that individuals
are as likely to organize such a concept around a prototype as they will a
natural category like bird or vegetable (seven turns out to be the prototypi-
cal odd number). The researchers concluded that Rosch has no way of
distinguishing between a classically defined and a natural concept. How-
ever, it seems to me that this line of evidence could as readily be cited as
showing that even "true" classical concepts have the kind of categorical
structure Rosch has ferreted out elsewhere. According to the latter analy-
sis, the tendency to discern naturalistic features in all categories is viewed
as a confirmation, rather than a refutation, of the natural view.
Another critical view, put forth by Daniel Osherson and Edward
Smith (1981), asserts that the prototype theory cannot account for the
347
I II I TowARD AN INTEGRATED CoGNITIVE SciENCE
ways in which more complex concepts (for example, the concept pet
fish) are apparently composed of elementary concepts. These authors pro-
pose a hybrid theory: that is, while categorization continues to need a core
concept, along the line of classical concept theory, this core aspect must
be combined with an identification process (the way in which one deter-
mines category membership). Rosch's approach is seen as an explanation
of how one identifies instances of the natural and artificial sort. In the view
of Osherson and Smith, such a hybrid theory is important because "the
ability to construct thoughts and complex concepts out of some basic stock
of concepts seems to lie near the heart of human mentation" (1981, p. 55).
To my mind, neither the findings of Armstrong and colleagues nor the
Osherson-Smith argument undermines the broad thrust of the Rosch posi-
tion. The classical theory remains feeble; the natural view emerges as a
more veridical description of how individuals form and utilize categories.
And yet the two lines of criticisms that I have reviewed do make an
important point: certain aspects of human cognition-such as the capacity
to apply the definition of odd number in a reliable manner or the ability to
form more elaborate concepts out of simpler ones-seem inexplicable in
terms of the natural view. At least some aspects of categorization may
employ the kind of computational operations classicists cherish.
348
A World Categorized
tion of color lexicons. Studying color lexicons from ninety-eight different
cultures, Berlin and Kay uncovered the following sequences. If (like the
Dani) a culture has only two color terms, terms will code for black and
white. If a third term is added, it will be red; fourth and fifth will be yellow
or green (followed by whichever one of the two has not been used); blue
and brown will be the next pair, competing for sixth and seventh places;
and purple, pink, orange, and gray will be the last four names to be coined.
These results are virtually impossible to account for through a series of
historical accidents or diffusions. Instead, the kinds of distinction cultures
make among colors, and elect to capture in their lexicons, apparently reflect
distinctions most salient in their perception of the world, presumably
(again) for neurophysiological reasons.
The Berlin and Kay findings were as epoch-making in the area of
anthropology as Rosch's were in psychology. Just as psychologists had to
reconsider their beliefs in classical definitions of concepts and in the arbi-
trariness of categories, anthropologists had to re-examine their beliefs in
the flexibility of naming and categorization schemes and in the influence
of language on thought structures. But some anthropologists feel that the
conclusions drawn by Berlin and Kay are too broad. John Bousfield, for
example, suggests that it is one thing to know that the color spectrum is
divided into two segments, but quite another to translate the terms as
"black" and "white." In his view, such a pair of terms would acquire far
more meaning (since they are the only ways to talk about colors) than they
do in a multicolor-term culture:
349
III I ToWARD AN INTEGRATED CoGNITIVE SciENCE
signify crucial cultural differences between life and death, noble and com-
mon, pure and impure. Even in our culture, terms like red or yellow, or blue or
green are appropriated to emphasize significant distinctions in "charged"
domains like politics, bodily states, religion. It is owing to their deep
concerns about phenomena like anger or patriotism or mourning-
phenomena that become associated with characteristic hues-that humans
attend to colors at all. To adopt this semantic point of departure, Sahlins
insists, is not to ignore biology but simply to assign it a proper place.
To my mind, it would be a mistake to get involved in a "chicken-egg"
question of which came first-the cultural distinctions motivating all hu-
mans or the ability to discriminate colors per se along certain lines. I read
Sahlins as making a different point. As he sees it, one does not confront
perception per se, except perhaps in an artificial situation where one asks
subjects to examine color chips. Perception is always marshaled in the
service of some cultural end. And so, a la Levi-Strauss, one ought to think
of the perceptual system-and globally, of the mind-as an implement of
culture, as an organizational structure to be exploited by the human cul-
tural enterprise. As Sahlins concludes, "It seems to me that Basic Color
Terms opens up very exciting prospects for an ethnography of color whose
general aim, quite beyond the determination of the empirical correlates of
semantic categories, might consist especially in the correlation of the semi-
otic and perceptual structures of color. For colors, too, are good to think
(with)" (1976, p. 16).
In such reactions to the Berlin and Kay findings, we see the same
impulse for territorial preservation which greeted the Rosch work. Just as
some psychologists felt the need to defend the classical view of concepts,
so, too, some anthropologists seek to ensure that the terrain of culture does
not disappear now that some potent universals have been demonstrated.
Perhaps scholars will eventually find some way of blending the strengths
of the two points of view, as Osherson and Smith attempt in psychology
and Sahlins attempts in anthropology. But to my mind, neither of these
lines of criticism undermines the main point of the natural view of catego-
ries. They are commentaries or cautionary notes, rather than convincing
revivals of the classical view in psychology or anthropology.
Just about the time that Eleanor Rosch had repaired to the wilds of New
Guinea, a shift in many ways parallel to the trends in psychology and
anthropology was coming to a head within the philosophical community.
350
A World Categorized
As has happened more than once, this philosophical shift was previewed
in the thinking of Ludwig Wittgenstein (1968). In the early part of the
century, as a principal contributor to, if not the inspirer of, the Vienna
circle (see pp. 62-65), Wittgenstein had stressed the importance of logic
and the need for precision in language (as well as the desirability of
keeping silent when such linguistic accuracy was not possible). But in the
second phase of his scholarly career, Wittgenstein came to focus on the
way in which ordinary language is used, and arrived at the controversial
conclusion that philosophical problems typically inhere in one's use of
language. He saw language as a loose and fragmentary set of elements and
as the necessary means of communication among individuals, but as prone
to obscure as well as to enlighten because it is the web through which all
other experience necessarily passes. Concepts are neither mental constructs
in the head nor abstract ideas in the world, but ought to be seen as
abilities which individuals can employ in ways acceptable to the rest of the
community-roughly speaking, as ways of accomplishing things.
In Wittgenstein's skeptical view, the most an analyst can hope to
achieve is greater insight into how language as a system works, and into
how our own ideas have come to be formed by the linguistic practices of
our community. Efforts to figure out what "really" happened, to sweep
aside the veil of language, are doomed to fail. And the glorification of logic,
or of abstract concepts bereft of utility within a community, is an irrele-
vant move, lacking philosophical force.
While Wittgenstein was calling for a radically different view of lan-
guage and conceptualization, the attack on logical empiricism was also
being pursued by once-sympathetic philosophers in the Anglo-American
tradition like W. V. 0. Quine (1953) and Nelson Goodman (1955). As I
noted in chapter 4, the distinction between the analytic and the synthetic,
the contrast between the immediate and the mediate, the possibility of
verifying observational sentences-these were all being spumed. More
modest goals for philosophy, or even the merging of philosophy with
psychology, as a kind of "naturalized epistemology," were being advanced.
Ultimately, a still more radical attack on the classical view of concepts
was put forth virtually in tandem by the American philosophers Saul
Kripke (1972) and Hilary Putnam (1975b ). This pair of scholars came to
question the belief that the world is a welter of sensation which can be
parsed with equal plausibility in an indefinite number of ways. Rather, in a
move analogous to the Roschian maneuver within psychology, these
philosophers embraced a view of naming and classifying that smacked of
the rankest realism. Kripke, Putnam, and their associates argued that there
is a real structure to the world, and that much of our conceptual armament is
designed to capture this genuine and attainable structure. Putnam (in his
talk of "natural kinds") and Kripke (in his talk of "rigid designators")
351
I I I I TOWARD AN INTEGRATED COGNITIVE SCIENCE
assumed that there are many objects, ranging from gold to goldfish, and also
many entities, like the persons "Richard Nixon" or "Greta Garbo," that
ought not be defined in terms of a list of criteria! attributes. For practical
purposes, these concepts have no definition, no intension; they have only an
extension, or a relation extending from the term to its concrete referent in
the world. Richard Nixon is the person whom his mother named as such; he
cannot be anyone else; and the only way to make sure that he is Richard
Nixon is to have a confirmed history that he is the same individual that his
parents so dubbed in the year 1913. By the same token, gold is what it is;
scientists may define it today as the atomic weight 79, but it will remain gold
even if the scientist's understanding of gold changes.
According to this view, all instances of the same natural kind possess
a definite underlying structural property in common. Thus, all lemons
ultimately possess the same structure even though any two randomly
selected lemons may lack a specific perceptible property in common. A
category is organized around stereotypes that enable a layperson to recog-
nize exemplars: such an organization proves essential inasmuch as one will
not necessarily be privy to the underlying genetic properties all lemons
really share.
As Stephen Schwartz (1979) has explained, we rely on experts to help
us to determine the appropriate use of such natural-kind terms. We assume
there is someone (or some procedure) that can confirm the identity of
Richard Nixon; we assume there is someone who can confirm that this
sphere is a lemon or that this metal is gold. Such natural-kind terms prove
susceptible to the formation of stable generalizations, such as, "If you do
X to gold, it will turn ...." To use the natural-kind term appropriately,
you do not need to know the trait governing the extension-indeed, at
present no one may know it; you just have to believe that it can be
discovered eventually. In other words, a nature is there (in principle) to be
discovered.
The question arises why we have natural-kind terms. Schwartz voices
the opinion that such terms turn up whenever the same stuff or thing
characteristically assumes a lot of different forms. Water can take different
forms, as can diseases, animals, plants, and so on. Seeing these changes
occurring on the "same" object, one assumes that there is some underlying
trait making the stuff continue to be of the same kind. In contrast, "artificial
kinds" do not necessarily exhibit dramatic changes: cars do not necessarily
rust, rings need not bend: these are simply vicissitudes to which any
man-made object may be subject. Schwartz goes on to discern a unity in the
changes exhibited by natural-kind terms, as all members of a particular kind
must go through the changes. Something of a law-governed nature must
remain the same in the transition from ice to water or from larva to butterfly.
352
A World Categorized
"Likewise the reason why I believe that tigers share an essence, the reason
why I am compelled to postulate an underlying trait for tigers that makes
tigers tigers, is that big tigers, by a perfectly natural, unconscious, and in-
eluctable process, produce little tigers" (1979, p. 315). Or to put it in Krip-
ke's terms, we come to think ofobjectsin termsofwhatcouldhappen to them,
in terms of the modal possibilities, the regularities and the exceptions.
To those somewhat removed from philosophical frays, the move to
natural-kind objects, the willingness to accept an ostensive definition,
("This is gold") or a history of occurrences ("This has always been
Nixon"), rather than a list of necessary and sufficient conditions ("Gold is
A,B,C, or Nixon is X, Y,Z) may seem innocuous enough. But within the
philosophical world, the moves first launched by Wittgenstein and then
extended in new and unanticipated directions by Kripke and Putnam are
considered very radical. Some sacred cows dating back to Aristotle are
directly challenged; some strong and biologically oriented contentions ("A
lemon has an essence ... ")are being introduced to a world that has been
comfortable in its empiricism and its nominalism. Rather than meanings
being located inside someone's head, the Putnam-Kripke tradition sug-
gests that meanings are located in the world-an imperative imposed by
the real structure of objects and the ways in which individuals come to
learn about them.
Discussion of the new theory of reference has been widespread within
the philosophical community. The talk of essences, of a "real" structure
in the world, of a connection between proper names (like Richard Nixon)
and natural-kind terms (like lemon or gold) is sometimes seen as a reac-
tionary move, one that recalls errors to which philosophers from a bygone
day were prey, but that had ostensibly been exorcised by several centuries
of empiricist philosophical analysis. The practice of collapsing diverse
elements into the same category because of ostensible underlying struc-
tural identities is seen as risky. Indeed, the reversion to realism has such
strong philosophical implications that even Putnam himself (though not
apparently Kripke) has had second thoughts about the dangerous lair into
which such an analytic line may lead.
Yet whatever misgivings an individual philosopher may feel, an ob-
server of the cognitive sciences is struck by the extent to which parallels
were occurring across several fields. The findings emerging empirically
from the laboratories of psychologists and from field trips to New Guinea
were closely reflecting the discussions engaged in by scholars who might
never think to conduct an experiment or join an anthropological team but
were deeply steeped in over two thousand years of argumentation on
related topics. From these widely different avenues came a common rejec-
tion of the classical view of concepts, a sharp impatience with criteriallists
353
I I I I TowARD AN INTEGRATED CoGNITivE SciENCE
of attributes, and a correlative belief that there are natural categories in the
world, best defined by examples (called variously "stereotypes" or "proto-
types"), and possibly reflecting underlying structures, which we as hu-
mans have the potential to recognize and to understand. In each case, the
extent to which the revisionist claims could take over the field remained
. in doubt, and some critics yearned for the good old classical days: yet the
challenge posed to the classical formulation was widely recognized and
frequently endorsed .
. · We have, then, a prototypic;al example of how several disciplines in the
cognitive sciences have combined to suggest a solution to a long-standing
philosophiCal puzzle. In this particular case, the puzzle surrounds the
human capacity to group together elements as members of a category, and
then to distinguish that category or class from others in the same general
domain of experience. A consensus had developed over the years that such
categorization was relatively arbitrary, with the culture determining which
entities ought to be grouped together. Most any set of criteria could be
proposed; and once proposed, these criteria could be applied reliably to
instances, as means of determining which belonged t0 the category and
which did not. Gasses were relatively clear; boundaries, relatively fixed.
Thus, it was appropriate for researchers like Jerome Bruner to conduct
experiments using artificially contrived classes: after all, they were thought
to represent well the kinds of classes we must utilize every day.
From the fields of psychology, anthropology, and philosophy, work-
ing initially in relative independence of one another, but possibly reflecting
the same factors or spirit, came an increasing realization that this classical
view is wrong. Categorization is anything but artificial and unambiguous;
rather, it relies on information in the natural world to which we as a species
are geared to respond. Categories have an internal structure, centered
around prototypes or stereotypes, with other instances being defined as
more or less peripheral depending upon the extent to which they share
pivotal features with the central prototype. Perceptual information is cru-
cial in defining the dimensions of a category; language, for the most part,
follows upon the discriminations made by the individuals, rather than
playing a controlling role in how one classifies in the world.
There is an irony in the decline of the classical theory-an irony we
have encountered in other contexts. The classical theory of concepts was
tailor-made to a computational model of mind-a set of precisely defined
dimensions which could, taken conjunctively, yield a category. Computers
were the perfect devices for simulating (or epitomizing) this mode of mind.
It is therefore striking that, at the very time that science had available a
device fully equal to simulating a certain view of concepts, that view was
found to be radically flawed. Not surprising, therefore, that there was
resistance across several disciplines to the relatively radical implications of
354
A World Categorized
the natural view of categorization. Many analysts hold out for a range of
categories, some of which are relatively natural, others of which are more
arbitrary and lend themselves readily to the kinds of analyses put forth in
the classical theory. And it is certainly true that the natural theory deals
better with colors or plants than with legal terms, religions, or self-con-
cepts; moreover the extent to which the natural theory handles man-made
artifacts satisfactorily is far from settled.
Indeed, to my mind, the principal challenge to the natural view of
categorization inheres in the extent to which it can move beyond an expla-
nation of those basic-level entities that we are "biologically prepared" to
perceive and classify. Color is most likely to lend itself to a natural-kind
account, because our perceptual system is primed to consider certain colors
as focal; and there is every reason to think that primates the world over
see colors in similar ways (unless they happen to be color-blind). The
evidence concerning other "natural" entities-such as plants, fruits, or
animals-is not as well worked out, but there is reason to expect that an
account in the natural-kind tradition may suffice (Berlin, Breedlove, and
Raven 1973).
But much of our classification in everyday experience extends beyond
the objects evolution has prepared us to classify in certain ways. There are
the multitude of man-made objects, ranging from tools to machines to
works of literature or art. There are more abstract concepts such as political
principles, religious precepts, belief systems, and economic laws. And there
are concepts that impinge much more upon an individual's own personal
concerns: concepts having to do with personality, motivation, sexuality,
emotions, and the self. Anthropologists in the ethnoscientific tradition have
attempted to apply to these concepts the approaches that worked in simpler
areas like kinship systems, plants, and colors. As I noted in chapter 8, these
attempts have not been notably successful; and, in fact, the discipline of
ethnoscience, in both its classical and its natural form, is quiescent at the
present time. One might say that the critique of the classical view has been
relatively successful but that the natural approach has not yet proved
genuinely illuminating with more complex, less "built-in" concepts.
Which leads to a more radical critique, of the sort associated with Clifford
Geertz of the Institute for Advanced Study, and other members of the
school of symbolic anthropology. Geertz (1973, 1983) opposes his position
355
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
356
A World Categorized
Of course, resolution of so fundamental a controversy cannot be pro-
vided in advance: it is, as they say, an empirical issue or perhaps a question
of which line of explanation will prove most satisfactory and revealing in
light of the canons judged to be relevant-be they canons of the natural
sciences, the social sciences, the humanities, or the arts. It is not, however,
simply a matter of taste.
In my own view, even if the processes by which individuals reason
or classify may be similar the world over, the actual products and the ways
they are thought about may be so different as to make illuminating gener-
alizations elusive. And to the extent that particular concepts are the result
of biases lurking within the linguistic system of a given culture, one risks
entering a Wittgensteinian-Whorfian morass from which few investigators
have yet returned. It seems to me that the important lines to be drawn are
between those concepts and those principles of reasoning that have
evolved since the early hominid era, as against those that have developed
only comparatively recently and reflect the particular history, customs, and
values of one or another social or cultural group. Oearly, there are ways
in which humans form concepts of colors or of animals, and these may well
be the same the world over. In formulaic terms, concepts may not be
formed classically, but a naturalistic approach to certain aspects of concep-
tualization and reasoning appears possible.
But when one enters the world that is entirely man-made, in arts,
rituals, or sciences, the possibility for effective generalization seems less
evident. Here, forces of history and culture may prove so dominant that
the differences may outweigh the similarities. While certain generaliza-
tions can be formed, the ones that seem most revealing about a particular
circumstance may not extend beyond that circumstance. What tells us the
most about the Balinese may tell us the least about the Belgians, and vice
versa.
Scientists cherish their own reflections and use them to guide their
studies. They devise artificial, criterially defined categories (else their work
could not begin), and they follow strict canons of logic (otherwise their
work would not be accepted). In their efforts to model thought processes,
they have sought such categories and such operations in normal individu-
als. But in their search for marks of the scientist in nonscientific man (or
even in the scientist, when not "playing" scientist), these investigators
have been disappointed. Faced with an artificial concept, the subject at-
tempts (if inappropriately) to treat it as if it were a natural concept. Asked
to reason logically, subjects have embraced images and stereotypes. So we
confront the paradoxical situation where scientists have attempted-for
experimental purposes-to make the natural world artificial; but their
subjects, faced with the difficulty of dealing with these artificial concepts,
357
I I I I TowARD AN INTEGRA TED CoGNITIVE SciENCE
358
A World Categorized
In confronting questions of categorization, we have moved into a
distinctly human terrain, involving both language and reasoning capaci-
ties. Few would quarrel with the claim that this area is less advanced than
the study of perception; indeed, as Geertz suggests, the possibility remains
that the most crucial questions about categorization may continue to elude
cognitive-scientific methods. Perhaps cognitive science as a field will prove
unequal to the task of explicating the genesis, nature, and use of our more
complex and evocative concepts and categories.
Yet, even if issues of categorization turn out to be susceptible to a
cognitive-scientific analysis; even if (say) the naturalistic view turns out
to be fully vindicated, there will remain crucial unresolved questions. For
classification or categorization is ultimately a tool-a means whereby in-
dividuals organize their world so as to solve certain problems and to
achieve certain ends. The ultimate achievements of this field will depend
on whether the principles evolved to deal with relatively simple concepts
and modest classification will prove applicable to complex concepts and
systems of concepts.
Just what sort of an organism will utilize these capacities, and toward
what ends will they be used? Here cognitive scientists have found them-
selves confronting value-laden aspects of human cognition-and, in par-
ticular, have directed considerable attention to the question whether in-
dividuals can be thought of as proceeding in a logical or rational fashion.
This question is of importance for scientific reasons, since much of cogni-
tive science is based on the model of the logical computer; it also has
important political and social implications. The answers to questions of
human rationality, with which I shall deal in the following chapter, turn
out to be surprising, and not particularly reassuring to those who place
their faith in reason.
359
13
How Rational a Being?
Whenever there is a simple error that most laymen fall for,
there is always a slightly more sophisticated version of the
same problem that experts fall for.
-AMos TVERSKY
While the Greeks sometimes defined man as a featherless biped, they had
far more invested in another definition: man as rational being. Only hu-
mans can invoke a form of thought that adheres explicitly to rules and
leads to conclusions that can be judged by the community as valid or
invalid. Unpacking this Greek notion a bit, it proves possible to distinguish
between the individual who reaches valid conclusions-and can therefore
be deemed rational-and the procedure by which these conclusions are
reached-a procedure termed logic. And so, one might conceive of a crea-
ture who is rational but reaches conclusions by means other than logic-
for example, by shrewd intuition, lucky guessing, or being programmed to
issue only valid responses.
Though surrounded (in a sense) by evidence of the irrationality of
human beings, philosophers have clung to the notion that human beings
are logical and rational-or at least to the ideal that human beings should
strive for rationality and that they have the potential to achieve it. This
preoccupation with an ideal of rationality is hardly surprising, since phi-
losophy itself has sought to proceed by rational means. Indeed, Bertrand
Russell once speculated that Aristotle was the first to define man in terms
of rationality (quoted in Pylyshyn 1984, p. 257). Over the years, the
development of the field of logic has been closely intertwined with the
history of philosophy: just as the logic of Aristotle's time informed Greek
philosophy so the logics of Frege, Whitehead and Russell, and Kripke have
informed the philosophy of today.
360
How Rational a Being?
Cognitive science was conceived in the shadow of contemporary logic.
As I noted in discussing the Hixon Symposium in chapter 2, both early
work on computers and the model of the neuron as a logical circuit encour-
aged a view of thought as logical (Jeffress 1951). Furthermore, the first
generation of cognitive scientists embraced a model of human beings that
was decidedly rationalistic. Jerome Bruner, Herbert Simon, Allen Newell,
and Jean Piaget all elected to investigate issues involving the abilities of
human beings to reason validly. In fact, the problems solved by the first
computer programs were problems in logic; the classification tasks investi-
gated by psychologists required logical deductive processes; Piaget went
one step further, not only studying problems of logic, but assuming that
"developed" humans reason by invoking principles of logic. As he once
put it, "reasoning is nothing more than the propositional calculus itself"
(Inhelder and Piaget 1958, p. 305).
But the faith of the initial generation in a study of logical problems
and its determined search for rational thought processes may have been
misguided. Empirical work on reasoning over the past thirty years has
severely challenged the notion that human beings-even sophisticated
ones-proceed in a rational manner, let alone that they invoke some logical
calculus in their reasoning. Once again, a computational age has docu-
mented departures from computerlike precision. This realization has
seeped through many lines of research, but few would quarrel with my
selection of Philip Johnson-Laird (from Great Britain) and the team of
Daniel Kahneman and Amos Tversky (originally from Israel) as the
spokesmen for this particular chapter of cognitive science. Not only have
these scholars developed stunning demonstrations of human departures
from rationality; they have also offered explanations of the reasons we
humans often go wrong in the ways we do.
361
I I I I TowARD AN INTEGRATED CoGNITivE SciENCE
it has an even number on the other." You are then allowed to turn over
two, but only two, cards in order to determine whether the rule is correct
as stated.
If you have already met this problem, or one of its close relatives, you
should have little trouble getting it correct. (In this case, your memory is
being tested, not your powers of reasoning.) But if this is your maiden
encounter, you will almost certainly miss it, as have over 90 percent of
subjects (including many logicians) to whom it has been presented in many
settings. Most subjects realize that there is no need to select the card
bearing the consonant, since it is clearly irrelevant to the rule; they also
appreciate that it is essential to turn over the card with the vowel, for an
odd number opposite would infirm the rule. The difficulty inheres in
deciding which of the two numbered cards to pick up. There is a strong
temptation to pick up the card with the even number, because the even
number is mentioned in the rule; and this temptation proves fatal to a
majority of subjects. But, in fact, it is irrelevant whether there is a vowel
or a consonant on the other side, since the rule does not actually take a
stand on what must be opposite to even numbers. On the other hand, it
is essential to pick up the card with the odd number on it. If that card has
a consonant on it, the result is irrelevant. If, however, the card has a vowel
on it, the rule in question has been infirmed, for the card must (according
to the rule) have an even (and not an odd number) on it.
The fact that this problem proves hard (even though, once explained, it
seems evident enough) gives one pause: ready explanations for this result
elude those who place great faith in the logical capabilities of their fellow
humans and themselves. But to my mind there is an even more interesting
twist to theW a son and Johnson-Laird demonstration. Consider the follow-
ing problem, again using four cards, each representing a journey. Each card
has a destination on one side and a mode of transportation on the other. This
time the cards have printed on them the legends, respectively, "Manches-
ter," "Sheffield," "Train," and "Car"; and the rule is: "Every time I go to
Manchester, I travel by train." While this rule is formally identical to the
number-letter version, it poses relatively little difficulty for individuals. In
fact 80 percent of subjects realize the need to turn over the card with the
word "car" on it. Apparently, one realizes that if the card with "car" on it
has the name "Manchester" on the back, the rule is infirmed; whereas it is
immaterial what it says on the back of the card with" train" on it since, as far
as the rule is concerned, one can go to Sheffield any way one wants.
Why is it that 80 percent of subjects get this problem correct, whereas
only one tenth know which cards to turn over in the logically identical
vowel-number version? Possibly a person is much better at solving a
problem entailing familiar material-material that enables one to place
oneself inside the situation, to figure out what one would do, what it
362
How Rational a Being?
would be reasonable to do. Indeed, you are more likely to succeed on the
aforementioned problem if you are British (and hence know of Sheffield
and Manchester) and if you are of an older generation (and so have grown
up with trains as a likely mode of travel).
The results clearly challenge any notion that individuals are "logic
machines," capable of applying the same modes of reasoning, independent
of the specific information in a problem. Thus, humans are notably different
from an ideal computer: the actual contents of a problem (vowels versus
vehicles) cannot make any difference for such a syntactically governed
apparatus. Yet just as the particular structure of each category colors a
subject's classifying strategies in a Roschian experiment, so, too, the content
of specific problems determines how all of us (except, perhaps, highly
trained logicians) will proceed on an apparently simple reasoning task.
Results on these and many other puzzles have convinced Philip John-
son-Laird, a leading cognitive scientist, that people do not employ a mental
logic in solving problems (1983). The kinds of logic described by logicians
simply seem irrelevant to normal individuals. We do not construct truth
tables and look up the result: we do not use formal rules of inference. But
Johnson-Laird maintains that there can be reasoning without logic. The
dilemma facing the cognitivist is to allow both for rationality and for
human error.
Puzzles about traveling or games with vowels and numbers may seem
a shade artificial, but there is nothing contrived about the syllogism. This
form of reasoning goes back at least to Greek times and was considered by
Aristotle (often our judge in these matters) as the core of logic. It is used
unreflectively by ordinary individuals as part of daily experience; and as
Johnson-Laird neatly shows, it is even drawn upon by critics who wish to
belittle the importance of syllogisms. One might say, "Syllogisms are
artificial." Furthermore, "Psychologists shouldn't study things that are
artificial." Therefore, "Psychologists should study the kinds of inferences
that are used regularly in daily life." In the very act of dismissal, this critic
has invoked syllogistic reasoning.
The subject (and anyone can play) is asked to figure out which conclu-
sions, if any, follow from these two premises. Johnson-Laird and his col-
363
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
leagues have found that such syllogisms pose considerable difficulties; and
that, often as not, people will draw invalid conclusions from them. More-
over, explaining the patterns of success and failure on such problems has
proved difficult for the logicians and psychologists who have sought a
comprehensive account. These misadventures have resulted because ex-
perts on syllogistic reasoning assume that individuals untrained in formal
logic nonetheless use some variant of it (for example, visual aids like Euler
circles or Venn diagrams) to solve the problems.
These experts have evolved elaborate accounts purporting to show
how the average person approaches syllogisms. These methods can per-
haps work in principle-if, for example, one has unlimited memory capac-
ity and has mastered sophisticated mathematical notations; but Johnson-
Laird illustrates convincingly that even the talented college student cannot
proceed by embracing these methods: "The present theories are too fragile
to bear the weight of human reason" (1983, p. 93). Either the theorists
succeed in accounting for deductive errors but fail to account for the
rationality that is exhibited by subjects, or they explain the ability of
subjects to reason adequately under ideal conditions but fail to illuminate
the kinds of error made by human subjects. The trick is to offer an explana-
tion that can account for both human rationality and human error. John-
son-Laird therefore introduces his major contribution to cognitive science:
the notion of a mental model.
He asks us to pretend that we have the power to conjure up individu-
als who fulfill one or more of the roles stated in the premises. Clearly, we
can create individuals who are at one time artists, beekeepers, and chemists
-who exhibit the trio of roles diSZussed in the premises. These individuals
can be represented in mental model form as
artist-beekeeper-chemist
artist-beekeeper-chemist
artist-beekeeper-chemist
We also know that there are individuals who are beekeepers and chemists
but not artists:
beekeeper-chemist
beekeeper-chemist
beekeeper-chemist
And we know from common sense (as well as from careful consideration
of the problem) that there can be chemists who are not beekeepers:
chemists
chemists
chemists
364
How Rational a Being?
We now have created a mental model that arrays all of the informa-
tion that can be taken directly from the premises given. We see that there
are three categories of individual: those who represent all three occupa-
tions; those who are beekeepers and chemists; and those who are just
chemists. (The fact that I have listed three instances in each category is
accidental; one may conjure up as many exemplars as one wishes.) Reading
off of these arrays, we can now proceed to answer questions or draw
conclusions. For example, if we want to determine whether all the artists
are also chemists, we can simply look at the tableaus and confirm that all
the artists are indeed chemists.
Johnson-Laird's approach is important in that it proceeds differently
from other tacks, which in some way follow the laws of formal logic, either
explicitly or implicitly. There is no need to translate the premises, implic-
itly or explicitly, into the p's and q 's of formal logic, to say that all p's are
q's, and all q's are rand to deduce, according to an algorithm, what follows
from that array. Nor is there any need to draw overlapping circles (the
visual embodiment of formal logic in a system like Euler circles) and to
read off which aspects overlap one another under each circumstance.
Rather, the individual engaged in constructing a mental model is simply
employing whatever medium is comfortable to him (words, images, some
hybrid) in order to represent the information for himself in a convenient
and readily accessible manner.
According to Johnson-Laird, one first represents the individual tok-
ens (artists, beekeepers) in some manner and thereby forms a mental
model of the first premise. Next, one adds the information in the second
premise (that beekeepers are chemists) to the mental model of the first
premise, taking into account the different ways in which this can be
done. An advisable strategy here is to establish the possible arrange-
ments (whether an individual can be a beekeeper without being an artist)
using as few "imaginary actors" as possible, though, as I have noted, the
precise number of actors depicted proves irrelevant to the syllogistic in-
ferences that are eventually drawn. Finally, after all of the mental pic-
tures have been constructed, an integrated set of pictures is submitted to
a test: a search is undertaken for an interpretation of the premises that is
inconsistent with the model. An inference is valid if, and only if, there is
no way of interpreting the premises that is consistent with a denial of
the conclusion.
In the example just given, it is possible to arrive at a valid conclusion
simply by constructing one mental model. Thus the solution of this prob-
lem is relatively simple: one avoids the explosion of combinatorial pos-
sibilities that plague conventional formal approaches to syllogistic reason-
ing. Of course, some other problems prove far more complex to model. For
example, Johnson-Laird introduced the following syllogistic premises:
365
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
If you believe that this syllogism is easy, you are almost certain to be
wrong. (Even if you know it is hard, you will probably fail.) In fact, as
Johnson-Laird shows, there is only one fool-proof conclusion:
All of these research maneuvers are carried out from the perspective
of psychology, though admittedly a psychology informed by a sophis-
ticated grasp of the ways and traps of logic. But Johnson-Laird penetrates
into the heartland of cognitive science because (like Stephen Kosslyn
working in the area of imagery) he has implemented his theory via a
computer program. The computer program consists of three basic steps:
366
How Rational a Being?
1. Construct a mental model of the first premise.
2. Add the information in the second premise to the mental model of the first
premise, taking into account the different ways in which this can be done. (This
turns out to be the trickiest aspect of the process, the one that determines whether
one, two, or three different models need be constructed.)
3. Frame a conclusion to express the relation, if any, between the "end"
terms that hold in all the models of the premises.
367
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
Since mental models can take many forms and serve many purposes, their
contents are very varied. They can contain nothing but tokens that represent
individuals and identities between them, as in the sorts of models that are required
for syllogistic reasoning. They can represent spatial relations between entities, and
the temporal or causal relations between events. A rich imaginary model of the
world can be used to compute the projective relations required for an image.
Models have a content and form that fits them to their purpose, whether it be to
explain, to predict, or to control. (1983, p. 410)
368
How Rational a Being1
guages; mental models which are structural analogues of the world; and
images that are the perceptual correlates of models from a particular point
of view (1983, p. 165). We ought to think of individuals as representing
information at several different levels of abstraction: moreover, the form
of representation at one level need not be the same as the form of represen-
tation at another level. Just as a computer may have several different
languages, ranging from machine code to a high-level programming lan-
guage, so, too, a psychological process might use only strings of symbols
at one level but involve images or mental models at a higher level of
representation. This ecumenical approach seems to do more justice to the
wide range of human psychological processes brought to bear on such
problems as text understanding or inference than does a stubborn adher-
ence to a single mode of mental representation.
Returning to the puzzle of logical reasoning, we find that Johnson-
Laird has enriched our understanding in two ways. To begin with, he has
shown that the ways in which human beings were once thought to ap-
proach problems of reasoning simply do not hold: one does not reason as
classical logic would suggest. Yet humans are not irrational either. Through
a careful analysis of syllogistic reasoning, the conduct of experiments with
subjects of different ages and degrees of expertise, and the simulation of
his model on a computer, Johnson-Laird has arrived at a picture of reason-
ing that seems robust and viable. A combination of several of the cognitive
sciences has yielded a genuine clarification of a long-standing philosophi-
cal issue-the degree of rationality of human beings.
Why does logic fail? As anthropologist Roy D' Andrade has noted, the
vocabulary of the logician is a second-order vocabulary (1982). It is a
statement not about things or events but, rather, about the consistency or
the inconsistency of statements. Ordinary language statements do not
usually refer to truth conditions but refer rather to states of affairs in the
world: people appear to be so designed (or so educated) that their major
interest focuses on what can happen in the world under such-and-such
conditions. The soundness, the speed, and the complexity of the reasoning
that individuals exhibit seem primarily a function of the degree of famil-
iarity and organization of the materials being processed, rather than a
function of any special or general ability of the person doing the reasoning.
And so there are appreciable differences in how a given person can reason
about different topics, topics that, from a formal point of view, call upon
the same degree (and even principles) of logical expertise.
This line of reasoning drawn from the analyses of Johnson-Laird,
Wason, and D'Andrade suggests that we can better understand the logical
reasoning of humans not by imputing to them any formal logical calculus
but by attending instead to two factors. The first has to do with content:
369
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
the greater the familiarity and the richer the relevant schemata which are
available, the more readily can one solve a problem. The second attribute
has to do with form: one succeeds on problems to the extent that one can
construct mental models that represent the relevant information in an
appropriate fashion and use these mental models flexibly. Just how one
learns to construct such mental models, to integrate them with "real
world" knowledge, and to deploy them appropriately in the proper
circumstances are fertile questions for developmental and educational
psychology.
If Johnson-Laird and company are correct, the kinds of principles
devised by logicians-and invoked by researchers like Piaget-will turn
out to have only limited applicability to how we reason in the real world.
Apparently we have evolved as creatures who are most likely to succeed
on tasks that contain familiar elements and allow the ready construction
and manipulation of mental models. Considerations of pure logic, a field
that developed long after our survival mechanisms had fallen into place,
may be useful for certain kinds of information under certain circumstances
by certain individuals. But logic cannot serve as a valid model of how most
individuals solve most problems most of the time.
370
How Rational a Being?
Theater Tickets and Coin Tossers
371
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
$10. Would you make the trip? Most people say that they will. Another
group is asked a similar question. This time the cost of the jacket is changed
to $15, and the cost of the calculator to $125 in the original store and to
$120 in the branch. Of respondents presented with this version, the major-
ity said that they would not make the extra trip. Note that in both cases
the total purchases are the same: the choice is always whether to drive
twenty minutes to save $5. But apparently respondents evaluate the saving
of $5 in relation to the price of the calculators. In relative terms, a reduction
from $15 to $10 (or 50 percent) is less resistible than a reduction from $125
to $120 (less than 5 percent).
Finally, meet Linda. She is thirty-one years old, single, outspoken, and
very bright. She majored in philosophy. As a student, she was deeply
concerned with issues of discrimination and social justice and also par-
ticipated in antinuclear demonstrations. You are now asked to rank, from
most to least probable, a series of eight statements. Included in the list are
the statements: "Linda is a psychiatric social worker," "Linda is a bank
teller" and "Linda is a bank teller and is active in the feminist movement."
On any rational account, it is more probable that Linda is a bank teller than
that Linda is both a bank teller and active in the feminist movement. The
probability of x, after all, is always greater than the probability of indepen-
dent event x and independent event y. Yet more than 80 percent of sub-
jects, including those who are sophisticated in statistics, assent more read-
ily to the statement that Linda is a bank teller and a feminist than to the
statement that Linda is a bank teller.
Why this clear flouting of rationality? Tversky and Kahneman see the
laws of probability being overwhelmed by the principle of representa-
tiveness. People are keenly attuned to the likelihood that someone with
certain characteristics will also exhibit other ones. (To the extent that
someone is a social activist, she is likely to be a feminist.) Indeed, the more
supporting details are consistent with this representative portrait, the more
likely a subject will assent to these new details, even though (on a rational
basis) the added details increasingly constrain the set and hence render it
less probable in absolute terms. Thus, given the information that Linda is
a certain kind of person, subjects readily fit in other events that have in
the past been representative of such persons, and in the process ignore
what they otherwise know about probability.
And this "otherwise" knowledge does still exist. Asked the abstract
question "Which is more probable, x alone or x and y ," subjects readily
consent that x alone is more probable. Moreover, when confronted with
the apparent contradiction between this abstract response and the Linda
question, they readily admit that they have made an error. Thus, it is too
simple to say that subjects were simply fooled: the response based on
372
How Rational a Being?
representativeness in the Linda problem seems to reflect a deep-seated bias
in human judgment.
These are but a few of the many vivid instances studied by Tversky
and Kahneman and, of the lot, among the easiest to present.* But these
examples should underscore the point that individuals do not reach deci-
sions in a way that is logically consistent or that obeys the laws of proba-
bility. I do not mean, however, that the behavior is illogical or inexplicable.
Rather, Tversky and Kahneman have revealed a separate psychology of
preferences which does not follow strictly from an economic calculation
of gains and losses but instead focuses on how individuals "frame" selec-
tions. Among the relevant considerations are: whether a person construes
the situation as one in which there is a guaranteed loss (as opposed to a
likely loss); whether a result will have a substantial effect on one's custom-
ary style of life; whether one has ready access to an instance of the category
in question; how one's imagination plays on what might have happened;
how closely an example resembles a prototype about which one already
holds strong views; the mood that one expects to be in; and the way in
which a question is actually phrased (for example, does one speak of the
killing of four hundred out of six hundred persons, or of the sparing of two
hundred out of six hundred?).
Overall, Tversky and Kahneman conclude that statistical principles
and rules of deduction are simply not imported from the kit of the mea-
surement scientist into the reasoning of everyday life. People are not
coherent in the way that it would be nice to believe they were. In even-
tempered tones, these investigators conclude, "The descriptive study of
preferences also presents challenges to the theory of rational choice be-
cause it is often far from clear whether the effects of decision weights,
reference points, framing, and regret should be treated as errors or biases
or whether they should be accepted as valid elements of human experi-
ence" (Kahneman and Tversky 1982, p. 171).
A Philosophical Critique
The results reported by Tversky and Kahneman, as well as those
gathered by Johnson-Laird, Wason, and other workers, have not gone
unnoticed ~r unre~arked upon by those keepers of human rationality-
the professzonal philosophers. One such is L. Jonathan Cohen of Oxford
University, who has taken on Tversky, Kahneman, and their colleagues in
a series of critiques. Cohen (1981) takes three swipes at the notion that
most individuals do not behave in a logical or rational fashion. The first
. •Interested re~ders can play subject by consulting the dozens of examples recounted
m Kahneman, Slov1c, and Tversky 1982.
373
III I TowARD AN INTEGRATED CoGNITIVE SciENCE
74
How RaHonal a Being?
and Tversky claim that this response reflects a serious error because it
indicates a failure to take into account the base rates-that is, the prior
probabilities. The investigators reason as follows. The fact that far more
cabs are blue than green must be taken into account in computing the
likelihood that the cab was, in fact, green. They then make four computa-
tions: the likelihood of correct identifications of the cab as blue, given that
it is blue (85% X .8 = 68% ); the incorrect identification of a cab as blue,
given that it is actually green (15% X .2 = 3% ); the correct identification
of a cab as green, given that it is actually green (15% X .8 = 12%); and
the incorrect identification of the cab as green, given that it is actually blue
(85% X .2 = 17% ). All together, the identifications of the cab as green
will be 29 percent (17% + 12%) and the fraction that is wrong will be
17/29. Consequently, according to the experimenters' mode of reasoning,
the probability that the cab in the accident was actually blue is 17/29 and
not 1/5.
But, argues Cohen, one need not approach this problem in this text-
book way (in this case via Bayes' theorem). He claims, for instance, that
jurors ought not to rely on probability if they can avoid doing so: such an
invoking of probability assumes that the issue before the court concerns
a long run of cab-color-identification problems, where actually the juror
is only trying to decide about this particular case. Strictly speaking, jurors
are here occupied with the probability that the cab actually involved in the
accident was blue, on the condition that the witness said it was green. Since
the jurors know that only 20 percent of the witnesses' statements about
cab colors are false, they rightly estimate the probability at issue as 1/5,
without violating Bayes' law. The fact that cab colors actually vary by an
85-to-15 ratio may properly be considered irrelevant to the estimate: after
all, this fact neither raises nor lowers the probability of a specific cab-color
identification being correct on the condition that it is an identification by
this particular witness. As Cohen phrases it, "A probability that holds
uniformly for each of a class of events because it is based on causal
properties, such as the physiology of vision, cannot be altered by facts,
such as chance distributions that have no causal efficacy in the individual
events" (1981, pp. 328-29).
Cohen comes up with another example that strikes much closer to
home. Suppose you are suffering from a disease which is either A or
B. It is known that A is nineteen times as common as B. The two diseases
are equally fatal if untreated, but it is dangerous to combine treatments.
Your physician orders a test which turns out to be correct on 80 percent
of the cases where a differential diagnosis must be made. The test reports
that you are suffering from the much rarer disease, B. Should you nonethe-
less opt for the treatment appropriate to A? According to the Bayesian
375
I I I 1 TowARD AN INTEGRATED CoGNITIVE SciENCE
view, this would be the proper step. In view of the prior distribution of
the disease, the probability that you are suffering from A will be 19/23.
But you might choose to ignore the prior probabilities, on the supposition
that the probability of your suffering from B is 4/5. According to Cohen,
it is irrational for you as a patient to answer in terms of the overall
probabilities rather than in terms of the particular diagnosis just made.
Paradoxically, if you followed the point of view in the statistical textbook,
it would be a waste of time even to apply the test for the disease, since
whatever its results, it is still more prudent on a statistical basis to assume
that you have disease A, the statistically more common one.
Cohen discerns different interests at stake here. The administrator
who wants to secure a high rate of diagnostic success for a particular
hospital at minimal cost would be well advised to follow the laws of
general probabilities, and so should eliminate the test. But the patient is
concerned with success in his own particular case, not with the probabilis-
tic success of the system in the long run, and so needs to evaluate test
results with respect to his chances. Long-run strings of success are of no
relevance to him.
In suggesting that the factors affecting the promulgators of a statistical
law are not identical to those influencing a solitary individual making a
decision about his or her own life, Cohen makes an interesting move. But
it has not won him much favor with other experts on judgment, such as
those asked to comment on his article in The Behavioral and Brain Sciences
(Cohen 1981). Cohen was castigated as an apologist for errors and for the
status quo, as a person who would undo the learning of science, and who
himself was incapable of putting forth an argument or reasoning coher-
ently and thus was dependent upon unsupported assertion and illusion.
These criticisms struck me as being excessively severe, as though Cohen
had hit a vulnerable spot. But to my mind, rather than demonstrating that
people are logical (as he has intended to do), Cohen has, in fact, provided
further evidence that they depart from logic and rationality in certain
systematic ways. He also has reminded us that acting in one's own ad-
judged self-interest is not necessarily the same as making a judgment
about which outcome is more likely to happen in some hypothetical event.
Some further points can be made as well. First of all, individuals have
evolved in order to survive in a certain range of biological and social
environments, not to satisfy some abstract notion of logic. It may well be
that the kinds of induction and deduction individuals habitually (and even
reflexively) make are most likely to lead to survival. And so they are
rational in this absolute sense, even if they appear irrational in the light
of certain textbook principles.
Another line of argument underscores the kinds of bias that have been
376
How Rational a Being?
built into favored psychological paradigms. To the extent that experiments
harbor illusions or tricks they may be seen as having limited relevance to
most daily behavior. Certainly if one slips traps into questions, it is easy
enough to secure wrong answers or answers that make subjects look fool-
ish. Piaget and his colleague Barbel Inhelder once designed a test in which
one gives young children a set of five tulips and two roses and then asks,
"Are there more tulips here or more flowers?" (Inhelder and Piaget 1964).
Until the age of seven, children will answer that there are more tulips,
seemingly oblivious to the fact that the term flowers subsumes all of the
tulips and all of the roses. Piaget and Inhelder confidently concluded that
young children are unable to compare a whole set (flowers) with one of
its subsets (the tulips). But as philosopher Jonathan Adler has pointed out
(1984), it may be too simple merely to assume that children are incapable
of making such comparisons. Rather, the children may simply not be
expecting the kind of question that violates ordinary conversational rules
of being "relevant" and pertinent. Children may just assume that the
comparison being asked for is the perceptually evident one-Are there
more tulips or more roses?-rather than the obscure if not downright
deceptive query, Are there more tulips or more tulips plus roses? On this
analysis, if children are asked for comparisons where a subclass could be
more legitimately compared with the entire class, they would answer cor-
rectly. And, indeed, it turns out that when such questions are posed to
children-Are there more children or more people?-they emerge as sensi-
tive to the principles of "class inclusion" (see also Donaldson 1978).
Applying this perspective to the preceding examples, we can see that
individuals might well be apprehending the Tversky-Kahneman problems
in light of the ways in which people usually converse and the kinds of
information they usually exchange. In the "Linda-bankteller" problem, for
example, one assumes that the information about Linda's previous political
involvements would only have been given if the listener was supposed to
draw the conclusion that she was a feminist. Thus, a listener simply makes
the reasonable assumption that the speaker wants him to believe that
Linda is a feminist--else why trouble to mention so much about her
political and social attitudes? In the problem about the taxicab, on the
other hand, it might be assumed that the information about base rates has
no relation to the question at hand. For as Tversky and Kahneman have
themselves pointed out, subjects respond differently when told, "Al-
though the two companies are roughly equal in size, 85 percent of the cab
accidents in the city involve green cabs and 15 percent involve blue cabs."
Now the information about base rates has been tied directly to the problem
at hand, and subjects do take it into account. It is too simple, I think, to
attribute subjects' apparent irrationality simply to the ways in which these
377
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
378
How Rational a Being?
they construe the study of rationality as a joint venture-one involving the
findings of empirical researchers as well as the clarifications and distinc-
tions of philosophical analysts. They leave open the possibility that in-
dividuals may depart from strict canons of logic and yet exhibit a viable
version of rationality. We see at work here a catholic model of cognitive
science. Instead of particular sciences banding together-as they have
done, say, in the study of psycholinguistics-there is a broader dialogue
at work-a dialogue between philosophers who are continuing to ponder
a classical philosophical problem, and empirical scientists who have de-
vised ingenious methods for attacking these venerable issues. And while
the examples here have been drawn largely from the realm of psychology,
they are by no means restricted to that discipline. Indeed, as I have sug-
gested, the kinds of concern explored by Wason and Johnson-Laird, by
Tversky and Kahneman, also involve issues of linguistics (How are the
particular questions phrased?); they can be applied to diverse anthropolog-
ical settings (How is reasoning carried out by different populations exhib-
iting different kinds of bias?); and they may be instructively simulated by
artificial intelligence (as Johnson-Laird has modeled his work on simula-
tion). Indeed, of the cognitive sciences, only the concerns of neuroscientists
stand relatively remote from work on logic, reasoning, and rationality.
Whether this statement is true only at present, or whether the whole realm
of rational and logical beliefs operates at a level apart from neuroscience,
has not yet been adequately considered.
Conclusion
379
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
beings typically form concepts of a classical sort and that they generally
reason in a logical fashion. Now the cognitive age, with its high-powered
computational techniques, has called into question the view of human
beings as operating in precise fashion. I do not mean, of course, that human
behavior is no longer subject to study by computational or other cognitive-
scientific techniques-indeed, Johnson-Laird has shown us just how some
such behavior can be accurately simulated; but the digital and deductive
fashion in which humans have been alleged to think is not viable. The
broader question remains whether various forms of human irrationality-
those documented by clinicians like Sigmund Freud and Carl Jung or by
anthropologists like Clifford Geertz and Dan Sperber-can be elucidated
by the methods of cognitive science.
Even negative lessons are important, and a science cannot be responsi-
ble for the message it yields, whether cheerful or gloomy. Yet challenging
the model of man-as-computer raises far-reaching questions of the extent
to which cognitive science has embraced the proper view of human menta-
tion and provided the proper methods for its study. I tum to these ques-
tions in the final chapter of the book, as there I revisit the major themes
of cognitive science in light of the histories I have related and the particular
lines of current research I have presented. In conclusion, I present my views
about the extent to which cognitive science has lived up to its initial
promise and delineate the principal paradoxes and challenges it confronts
at the present time.
380
14
Conclusion:
The Computational
Paradox and the
Cognitive Challenge
Surveying the scientific landscape at the beginning of the century, a far-
sighted observer might have felt justified in announcing the arrival of the
mind's new science. After all, building on the philosophical tradition of the
Greeks and the Enlightenment, and in the wake of dramatic breakthroughs
in physics, chemistry, and biology, the solution to the mystery of human
mental processes seemed at hand. Moreover, toward the end of the nine-
teenth century, a raft of new disciplines concerned particularly with
human thought and behavior had been launched. Surely the opportunity
to look at individuals in many cultures, in the light of the latest findings
about the human nervous system and with the powerful tools of logic and
mathematics, should sooner or later yield a bona fide science of the mind.
From a contemporary perspective, it seems evident that at least three
conditions had to fall into place before this dream could reach fruition.
First of all, it was necessary to demonstrate the inadequacies of the behav-
iorist approach. Second, the particular limitations of each social science had
to be acknowledged. Finally, the advent of the computer was needed to
provide the final impetus for a new cognitive science.
381
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
382
Conclusion: The ComputaHonal Paradox and the CogniHve Challenge
past few decades and to discern what remains to be accomplished if cogni-
tive science is to achieve its full potential. This evaluation will entail a
consideration of the central concept in cognitive science-that of the repre-
sentational level-as well as a re-examination of two themes introduced
in the opening chapters of this book, the computational paradox and the
cognitive challenge.
383
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
384
Conclusion: The CompufaH.onal Paradox and the Cognitive Challenge
the emergence of the computer in our time. Since the first generation of
cognitive scientists, the computer has served as the most available and the
most appropriate model for thinking about thinking. And for most, it soon
became indispensable in their daily empirical and theoretical work.
Though the linking of computation and cognitivism turns out to have been
a contingent rather than a necessary fact, the fate of cognitive science is
closely tied to the fate of the computer.
And this leads to that strange state of affairs I have dubbed the
computational paradox. With the vigorous tradition, since the time of the
Greeks, of thinking about human thought as an embodiment of mathemat-
ical principles, it is hardly surprising that the first generation of cognitivists
-reared in the logical positivist tradition-should have embraced a highly
rationalistic view of human thought. One of the major results of the first
years of cognitive science, however, has been a challenge of that ready
assumption.
To be sure, when it comes to elementary and relatively "impenetra-
ble" processes like visual perception or syntactic analysis, an authoritative
computational account may some day be given. That is, the kinds of
descriptions that are legitimately offered in the terms of a digital von
Neumann computer may tum out to be appropriate accounts of these
human cognitive processes as well. But as one moves to more complex and
belief-tainted processes such as classification of ontological domains or
judgments concerning rival courses of action, the computational model
becomes less adequate. Human beings apparently do not approach these
tasks in a manner that can be characterized as logical or rational or that
entail step-by-step symbolic processing. Rather, they employ heuristics,
strategies, biases, images, and other vague and approximate approaches.
The kinds of symbol-manipulation models invoked by Newell, Simon, and
others in the first generation of cognitivists do not seem optimal for de-
scribing such central human capacities.
The paradox lies in the fact that these insights came about largely
through attempts to use computational models and modelingi only
through scrupulous adherence to computational thinking could scientists
discover the ways in which humans actually differ from the serial digital
computer-the von Neumann computer, the model that dominated the
thinking of the first generation of cognitive scientists.
I must again underscore one point. By insisting on the computational
paradox, I do not mean to assert that it is impossible to arrive at a computa-
tional account of human behavioral and thought patterns in all of their
perversity, irrationality, and subjectivity. Such accounts may well be pos-
sible and-as has been known since the time of Turing-are certainly
possible in principle. Rather, the paradox suggests that the portrait of
385
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
386
Conclusion: The Computational Paradox and the Cognitive Challenge
ing a priori to what extent the ways currently embraced for describing the
representations of computers will prove germane to organisms, be they
paramecia or professors. The kinds of representations favored by neo-
associationists like Geoffrey Hinton tum out to be radically different
(and much sparer) than those countenanced by Jerry Fodor or Zenon
Pylyshyn; moreover, it may turn out that neither is adequate or suited
for describing an individual who is dreaming, writing a poem, or listen-
ing to music. Earlier models of thought-the reflex arc, the hydraulic
engine, the telephone switchboard-are now seen to be extremely lim-
ited. It is already clear that one kind of computer does not suffice to
model all thought. We must face the alternative that humans may be an
amalgam of several kinds of computers, or computer models, or may de-
viate from any kind of computer yet described. Computers will be pivo-
tal in helping us determine how computerlike we are, but the ultimate
verdict may be "Not very much."
Even if computers emerge as viable models for certain facets of human
thought, the question arises about the various aspects of human nature
that have been bracketed by cognitive scientists. As I noted in chapter 3,
nearly all cognitive scientists have conspired to exclude from consideration
such nontrivial factors as the role of the surrounding context, the affective
aspects of experience, and the effects of cultural and historical factors on
human behavior and thought (see D. Norman 1980). Some take the posi-
tion that this is only a temporary move, until the relatively discrete aspects
of cognition have been unraveled; others take the stronger positions that
cognitive science should never deal with these aspects or even that a
cognitive-scientific account will ultimately render unnecessary any ac-
count of these "fuzzier" factors.
Even a brief consideration of each of these "bracketed" topics would
require many pages, and since cognitive scientists have themselves steered
clear of these issues, there is little work within the disciplinary tradition
on which I can draw. My own belief is that, ultimately, cognitive science
will have to deal with these factors in one of two ways. Either scientists
will propose a cognitive account of affect in which, for example, affective
states will be viewed as quantitative values along a dimension, like happi-
ness or cruelty; or researchers will opt for a complex explanatory frame-
work in which the interaction of traditional cognitive factors with affective
or cultural factors can somehow be modeled. These will be important but
enormously difficult undertakings, for which traditional computational
considerations may provide scant help.
Scholars differ widely from one another in their intuitions about the
extent to which these other factors may ultimately engulf cognitive fac-
tors. From the perspective of a philosopher like Hubert Dreyfus, a linguist
387
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
388
Conclusion: The Computational Paradox and the Cognitive Challenge
The Cognitive Challenge
389
I I I I TowARD AN INTEGRATED CoGNITIVE SciENCE
ing cognitive scientists"; otherwise, they are simply doing their own
thing.
This "weak" version of cognitive science is quite possibly the norm
today but scarcely warrants the label of an important new science. In a
stronger, more gritty version of cognitive science, there will be gradual
attenuation of disciplinary boundaries and loyalties. These will be replaced
by a concerted effort by scientists committed to a representational account
to model and explain the most crucial human cognitive functions.
This reconfiguration of the territory of cognitive science rests on the
following analysis. Today, what is most central to cognitive studies is an
individual's disciplinary background: whether one works as a philosopher
or an anthropologist is more salient than whether one works on issues of
language or of social interaction. This organization around the traditional
disciplines would be appropriate if the actual domains of cognition did not
make a central difference; so long as the same processes are believed to
occur irrespective of the content of a domain (musical versus spatial cogni-
tion, for example), the conventional disciplinary division of labor makes
sense.
I hold a very different, and still controversial, vision. From my per-
spective, as elaborated in this book, the crucial divisions within cognitive
science are not the traditional disciplinary perspectives but rather the spe-
cific cognitive contents. Therefore, scientists should be characterized by
the central cognitive domain on which they work: broad domains like
language, music, social knowledge, logical thought; and more focused sub-
domains, like syntactic processing, the early phases of visual processing,
or the perception of rhythm. Scientific training and research enterprises
should come increasingly to be organized around these problems. When
working on these problems, scientists should fuse their necessarily differ-
ent perspectives in order to arrive at a full account of the particular cogni-
tive domain at issue. And so the ultimate cognitive-scientific picture of
syntactic processing, or of language as a whole, should be a coordinated
representational account which covers the full gamut of the traditional
disciplines without any need even to mention them.
Yet the question of disciplines or, more broadly, of levels of explana-
tion cannot be bypassed entirely-and here we confront the major chal-
lenge to contemporary cognitive studies. Having established the legitimacy
of the representational level, cognitive workers must trace out the ways in
which this level maps onto the other legitimate (and legitimized) ways in
which human activities can be construed. For a time, believers in the
representational level had to proceed along their own, as yet unexplored
path-and adherence to this single-minded program was the genius of the
pioneering generation of the 1950s. But ultimately, such splendid isolation
must be shattered. We must come to understand how culture is mapped
390
Conclusion: The Computational Paradox and the Cognitive Challenge
onto brains-and the royal road toward such understanding will be the
representational level.
The reason for such linkage across levels is simple but crucial. Unless
the significance of work in each science can be connected to that under-
taken in neighboring areas, the significance (and the limitations) of that
work cannot be appreciated. No one fears the demise of physics, chemis-
try, and biology; and yet each of these discipines has vital, articulated,
necessary links to the next level, through "borderland" disciplines like
physical chemistry or biochemistry.
But, paradoxically, much of the best work in cognitive science has
been carried out as if only the level of mental representation existed. In the
case of language (more specifically, grammar), for example, the brilliant
work of Chomsky and his followers makes no reference to, and could be
maintained irrespective of, the actual conditions in the brain and in the
surrounding culture. If cognitive science is to mature, however, the ulti-
mate representational account of language must relate, at one extreme, to
knowledge about the neural architecture of certain regions of the left
hemisphere of the brain; and, on the other, to knowledge about the struc-
ture and function of language in different cultural groups. Only such a
linking of levels can indicate whether proposed representational accounts
of language are in fact appropriate, in light of neural and cultural consider-
ations. The goal of this penetration of levels is not, to repeat, so that one
discipline or level can swallow the other; but rather, so that our under-
standing of a domain like language can touch on all the relevant scientific
perspectives, from neuron to nation.
Just as many committed cognitive scientists have restricted their work
to the representational level and have spumed the borderland territory, so
too they have called for a narrow delineation of what counts as cognitive.
Thus, Jerry Fodor (1981) has expressed skepticism about the capacity of
cognitive science to explain any of the higher or inore complex forms of
thought, which are "permeable" to a person's beliefs; and Zenon Pylyshyn
(1984) has proposed a definition of cognition which excludes areas like
learning, development, and "moods."
Although perhaps a prudent research strategy in the short term, I find
this a misguided overall program for the cognitive sciences. Just because
our current measures or concepts are primitive, we ought not to violate a
common-sense notion of what mind is about; even less should we want
to bypass the most impressive achievements of the human mind.
Indeed, in my view, the ultimate goal of cognitive science should be
-precisely-to provide a cogent scientific account of how human beings
achieve their most remarkable symbolic products: how we come to com-
pose symphonies, write poems, invent machines (including computers), or
construct theories (including cognitive-scientific ones). Such accounts will
391
I I I I TOWARD AN INTEGRA TED CoGNITIVE SciENCE
392
REFERENCES
393
References
Berlin, B.; and Kay, P. 1969. BflSic Color Terms: Their UniversRiity Rnd Evolution. Berkeley: Univer-
sity of California Press.
Bernstein, J. 1981. "Profiles: Marvin Minsky and Artificial Intelligence," New Yorker 57:
50-126.
- - - . 1982. Science Observed: EsSRys Out of My Mind. New York: Basic Books.
Berwick, R.; and Weinberg, A. 1983. "The Role of Grammars in Models of Language Use,"
Cognition 13: 1-61.
Block, N., ed. 1981R. lmRgery. Cambridge, Mass.: MIT Press, Bradford Books.
- - - , ed. 1981b. Readings in Philosophy of Psychology, vol. II. Cambridge, Mass.: Harvard Univer-
sity Press.
- - - . 1983. "Mental Pictures and Cognitive Science," The PhilosophicAl Rroiew 92: 499-541.
Bloomfield, L. 1925. "Why a Linguistics Society?," LRnguRge 1: 1-5.
- - - . 1933. LRnguage. 2nd. ed. New York: Holt, Rinehart & Winston.
- - - . 1936. "Language or Ideas?," LRnguage 12: 89-95.
- - - . 1943. "Obituary of Franz Boas," LRnguage 19: 198.
Blumenthal, A. L. 1979. "Psychology and Linguistics: The First Half-century." Lecture deliv-
ered at the APA centennial symposium: "Linguistics and Psychology," New York.
Bobrow, D. G. 1968. "Natural Language Input for a Computer Problem-solving System." In
M. L. Minsky, ed., SemRntic Information Processing. Cambridge, Mass.: MIT Press.
---;and Winograd, T. 1977. "An Overview of KRL, a Knowledge Representation Lan-
guage," Cogniti'OI! Science 1: 3--46.
Boden, M. 1977. Arti/iciRllntelligence and NRiurRI MRn. New York: Basic Books.
- - - . 1981. Minds Rnd MechRnisms. Ithaca, N.Y.: Cornell University Press.
Boffey, P.M. 1983. "'Rational' Decisions Prove Not to Be," New York Times, 6 December 1983,
pp. Cl, C7.
Bolinger, D. 1968. Aspects of LRnguRge. New York: Harcourt, Brace & World.
Boring, E. G. 1950. A History of ExperimentAl Psychology. New York: Appleton-Century-Crofts.
Bousfield, J. 1979. "The World Seen as a Colour Chart." In R. F. Ellen and D. Reason, eds.,
ClRssi/icRtions in Their SocUll Context. London: Academic Press.
Bower, G. 1976. "Comprehending and Recalling Stories." Address to the American Psycho-
logical Association, Washington, D.C.
Brainerd, C. J. 1978. "The Stage Question in Cognitive-Developmental Theory," The Behavioral
Rnd BrRin Sciences 2: 173-213.
Bransford, J. D.; and Franks, ]. J. 1971. "The Abstraction of Linguistic Ideas," Cogniti'OI!
Psychology 2: 331-50.
Bransford, J. D.; Franks, J. J.; Morris, C. D.; and Stein, B. S. 1979. "Some Constraints on
Learning and Memory Research." In L. S. Cermak and F. I. M. Craik, eds., Lroels of
Processing Rnd HumRn Memory. Hillsdale, N.J.: Erlbaum.
Bransford, }. D.; and Johnson, M. V. 1972. "Contextual Prerequisites for Understanding:
Some Investigations of Comprehension and Recall," Journal of VerbRI Learning Rnd VerbRI
BehRvior 11: 717-21.
Bransford, J.D.; and McCarrell, N. S. 1975. "A Sketch of a Cognitive Approach to Compre-
hension: Some Thoughts about Understanding What It Means to Comprehend." In P.
N. Johnson-Laird and P. C. Wason, eds., 1977. Thinl:i.ng: Readings in Cognitive Science. Cam-
bridge: Cambridge University Press.
Brentano, F. 1874. Psychologie vom empirischen StRndpunkte. Leipzig.
Bresnan, J, 1978. "A Realistic Transformational Grammar." In M. Halle, et al., eds., Linguistic
Theory Rnd PsychologicRI ReAlity. Cambridge, Mass.: MIT Press.
- - - . 1981. "An Approach to Universal Grammar and the Mental Representation of Lan-
guage," Cognition 10: 39-52.
Brew, J. 0., ed. 1968. One Hundred Years of Anthropology. Cambridge, Mass.: Harvard University
Press.
Brewster, D. 1854. North British Rroiew. Quoted in J. Miller, 1983. States of Mind. New York:
Pantheon Books.
Broadbent, D. E. 1954. "The Role of Auditory Localization in Attention and Memory Span,"
fournRI of ExperimentAl Psychology 47: 191-96.
- - - . 1980. "The Minimization of Models." In A.}. Chapman and D. M. Jones, eds., Models
of MRn, pp. 113-27. London: British Psychological Society.
Broca, P. 1861. "Remarques sur Ia siege de la faculte du langage articule," Bulletin de Ia societe
d'Rnlhropologie (Paris) 6.
394
References
Brown, C. M. 1984. "Computer Vision and Natural Constraints," Science 224: 1299-1305.
Brown, R. 1956. "Language and Categories." In J. S. Bruner et al., A Study of Thinking. New
York: John Wiley.
- - - . 1975. "Reference: In Memorial Tribute to Eric Lenneberg," Cognition 4: 125-53.
- - - ; and Lenneberg, E. H. 1954. "A Study in Language and Cognition," Journal of Abnormal
and Social Psychology 44: 454--62.
Bruner, J. S. 1944. Mandate from the People. New York: Duell, Sloan & Pearce.
- - - . 1973. Beyond the Information Given; Studies in the Psychology of Knowing. ]. Anglin, ed. New
York: W. W. Norton.
- - - . 1982. In Search of Mind. Preliminary version.
- - - . 1983. In Search of Mind. New York: Harper & Row.
--;Goodnow,].; and Austin, G. 1956. A Study of Thinking. New York: John Wiley.
Carnap, R. 1959. "Psychology in Physical Language." In A. J. Ayer, ed., Logical Positivism.
Glencoe, Ill.: Free Press.
- - - . 1967. The Logical Structure of the World; Pseudoproblems in Philosophy. Berkeley: University
of California Press. Original work published in 1928.
Cazeneuve, J. 1968. "Levy-Bruhl, Lucien." Entry in International Encyclopedia of the Social
Sciences, ed. by D. Sills. New York: Macmillan.
- - - . 1972. Luden Uvy-Bruhl, trans. by P. Riviere. Oxford: Basil Blackwell.
Chemiak, C. 1983. "The 'universal acceptance of logic.'" Berkeley Cogniti"OI! Science Report, no. 6.
Cherry, E. C. 1953. "Some Experiments on the Recognition of Speech, with One and with
Two Ears," Journal of the Acoustical Sodety of America 25: 975-79.
Chomsky, N. 1955. "The Logical Structure of Linguistic Theory." Doctoral dissertation,
University of Pennsylvania. Published as a monograph by Plenum Press, New York, in
1975.
- - - . 1957. SyntRCtic Structures. The Hague: Mouton.
- - - . 1964R. "A Review of B. F. Skinner's Verbal Behavior." In J. A. Fodor and J. J. Katz, eds.,
The Structure of LRnguRge: Readings in the Philosophy of LRnguRge, pp. 547-78. Englewood Cliffs,
N.J.: Prentice-Hall. Original work published in Language 35 (1959): 26-58.
- - - . 1964b. "A Transformational Approach to Syntax." In J. A. Fodor and J. }. Katz, eds.,
The Structure of LRnguRge: ReRdings in the Philosophy of Language, 1964, pp. 211-45. (Reprinted
from A. A. Hill, ed. (1962). Proceedings of the Third TaRS Conference on Problems of Linguistic
Analysis in English, 1958, pp. 124-58. Austin, Tex.: University of Texas Press.) Englewood
Cliffs, N.J.: Prentice-Hall.
- - - . 1965. Aspects of the Theory of Syntax. Cambridge, Mass.: MIT Press.
- - - . 1966. Cartesian Linguistics. New York: Harper & Row.
- - - . 1972. LanguRge and Mind. New York: Harcourt Brace Jovanovich.
- - - . 1975. Reflections on LRnguage. New York: Pantheon.
- - - . 1979. Language and Responsibility. New York: Pantheon. Originally published in French,
1977.
- - - . 1980. Rules and Representations. New York: Columbia University Press.
- - - . 1981. "Knowledge of Language: Its Elements and Origins," Philosuphical TransRCtions of
the Royal Society of London, Series B, 295: 223-34.
- - - . 1982. Some Concepts and Consequences of the Theory of Government and Binding. Cambridge,
Mass.: MIT Press.
- - - . 1984. "Changing Perspectives on Knowledge and Use of Language." Paper delivered
at the Society for Philosophy and Psychology, Massachusetts Institute of Technology,
Cambridge.
---;and Halle, M. 1968. The Sound Pattern of English. New York: Harper & Row.
Clowes, M. 1971. "On Seeing Things," Arti/icial Intelligence 2: 79-116.
Cohen, G. 1977. The Psychology of Cognition. London, New York: Academic Press.
Cohen, L. J. 1981. "Can Human Irrationality Be Experimentally Demonstrated?," Behavioral
and Brain Sciences 4: 317-70.
Colby, K. M. 1975. Artificial Paranoia. New York: Pergamon Press.
Cole, M.; and Scribner, S. 1974. Culture and Thought. New York: John Wiley.
Collins, A. 1977. "Why Cognitive Science," Cogniti"OI! Science 1: 3-4.
Copleston, F. 1964. A HIStory of Philosophy: Modern Philosophy: The British Philosophers. New York:
Doubleday.
Craik, F. I. M.; and Lockhart, R. S. 1972. "Levels of Processing: A Framework for Memory
Research," Journal of Verbal leRrning and Verbal Behavior 11: 671-84.
395
References
Crick, M. 1976, Explorations in LRnguage and MNning: ToWRrds a Semantic Anthropology. New York:
John Wiley.
D' Andrade, R. G. 1981. "The Cultural Part of Cognition," Cognitive Science 5: 179-95.
- - - . 1982. "Reason versus Logic." The Ecology of Cognition: Biological, Cultural and Historical
Perspectives. Symposium, Greensboro, N.C.
Dasen, P.R. 1972. "Cross-Cultural Piagetian Research: A Summary," jOJtmRl of Cross-Cultural
Psychology 3: 23-24.
---;and Heron, A. 1981. "Cross-Cultural Tests of Piaget's Theory." In H. C. Triandis and
A. Heron, eds., Handbook of Cross-Cultural Psychology, vol. 4: Developmenflll Psychology. Boston:
Allyn & Bacon.
Davis, M. 1958. Computability and UnsolMbility. New York: McGraw-Hill.
Dejerine, J. 1892. "Contribution a !'etude anatomo-pathologique et clinique des differents
varietes de cecite verbale." Comp. rend. sceRn. soc. bioi. 9( 4): 61-90.
Dennett, D. C. 1978. Brainstorms: Philosophical Es5Rys on Mind and Psychology. Cambridge, Mass.:
MIT/Bradford Books.
- - . 1983. "Cognitive Wheels: The Frame Problem of A.I." Unpublished paper, Tufts
University.
Dennis, M. 1980. "Language Acquisition in a Single Hemisphere: Semantic Organization."
In D. Caplan, ed., Bwlogical Studies of Mental Processes. Cambridge, Mass.: MIT Press.
- - - ; and Whitaker, H. A. 1976. "Language Acquisition Following Hemidecortication:
Linguistic Superiority of the Left over the Right Hemisphere," Brain and LRnguage 3:
404-33.
Descartes, R. 1951. Meditation on First Philosophy, trans. by L. J. Lafleur. New York: Library on
Liberal Arts: Liberal Arts Press. Original work published in 1641.
de Valois, R. L.; and Jacobs, G. H. 1968. "Primate Color Vision," Science 162: 533-40.
De Waal Malefijt, A. 1974./mt~ges of Man: A History of Anthropologic11l Thought. New York: Alfred
A. Knopf.
Diamond, S., ed. 1974. The Roots of Psychology. New York: Basic Books.
Dodwell, P. 1971. "Is a Theory of Conceptual Development Necessary?" InT. Mischel, ed.,
Cognitive Development 11nd Epistemology. New York: Academic Press.
s
Donaldson, M. 1978. Children Minds. London: Fontana.
Danders, F. C. 1868. "On the Speed of Mental Processes." Reprinted in Ada Psychologica 30
(1969): 412-31.
Dorf, R. C. 1974. Computers 11nd M11n. San Francisco: Boyd & Frazer.
Dresher, B. E.; and Hornstein, N. 1976. "On Some Supposed Contributions of Artificial
Intelligence to the Scientific Study of Language," Cognition 4: 321-48.
Dreyfus, H. 1972. What Computers Can't Do: A Critique of Arti/icilll ReRSon. New York: Harper &
Row.
Duncker, K. 1945. "On Problem Solving," Psychological Monographs 58 (whole no. 270): 5.
Durant, W. 1926. The Story of Philosophy. New York: Simon & Schuster.
Durbin, M. 1966. "The Goals of Ethnoscience," Anthropological LingJtistics 8: 22-33.
Ebbinghaus, H. 1913. Memory: A Contribution fo Experimental Psychology. New York: Columbia
Teacher's College. Originally published 1885.
Edgerton, R. B.; and Langness, L. L. 1974. Methods 11nd Styles in the Study of Culture. New York:
Thomas Crowell.
Ehrenfels, C. von. 1890. "Uber Gestaltqualitaten." Vtijsch. Wrss. Philos. 14: 249-92.
Evans, T. G. 1968. "A Program for the Solution of Geometric-Analogy Intelligence Test
Questions." In M. L. Minsky, ed., Semantic lnfarmation Processing. Cambridge, Mass.: MIT
Press.
Fahlman, S.C.; Hinton, G. E.; and Sejnowski, T. J. 1983. "Massively Parallel Architecture for
AI; NETI, Thistle and Boltzmann Machines." Unpublished manuscript, Carnegie-Mel-
lon University.
Fancher, R. E. 1979. Pioneers of Psychology. New York: W. W. Norton.
Farah, M. J. 1984. "The Neurological Basis of Mental Imagery; A Componentical Analysis,"
Cognition 18: 243-61.
Fechner, G. T. 1912. "Elemente der psychophysik." In B. Rand, ed., Classical Psychologists, trans.
by H. S. Langfeld. Boston: Houghton-Mifflin. Original work published in 1860.
Feigenbaum, E. A. 1959. "An Information-Processing Theory of Verbal Learning." Doctoral
dissertation, Pittsburgh: Carnegie Institute of Technology, 1959. Rand Corporation
Paper, P-1817.
396
References
---;Buchanan, B. G.; and Lederberg, J. 1971. "On Generality and Problem Solving: A Case
Study Using the DENDRAL Program." In B. Meltzer and D. Michie, eds., Machine
Intelligence, vol. 6. Edinburgh: Edinburgh University Press.
Feigenbaum, E. A.; and Feldman, J., eds. 1963. Comflltters and Thought. New York: McGraw-Hill.
Feigenbaum, E. A.; and McCorduck, P. 1983. The Fifth Genm1tion: Artificial Intelligence and Japans
Comflltter Challenge to the World. Reading, Mass.: Addison-Wesley.
Feldman, J. A. 1981. "A Connectionist Model of Visual Memory." In G. E. Hinton and J. A.
Anderson, eds., Parallel Models of Assodati'OI! Memory. Hillsdale, N.J.: Lawrence Erlbaum.
Finke, R. A.; and Schmidt, M. J, 1978. "The Quantitative Measure of Pattern Representation
in Images using Orientation-specific Color After-effects," Perception and Psychophysics 23:
515-20.
Fodor, ]. A. 1972. "Some Reflections on L. S. Vygotsky's 'Thought and Language,'" Cog-
nition 7: 83-95.
- - - . 1975. The language of Thought. New York: Thomas Y. Crowell.
- - - . 1980. "Methodological Solipsism Considered as a Research Strategy in Cognitive
Psychology," Behavioral and Brain Scienm 3: 63-110.
- - . 1981a. "The Mind-Body Problem," Scientific American 244: 114-23.
- - - . 1981b. Representations: Philosophical Essays on the Foundations of Cogniti'OI! Science. Cambridge,
Mass.: MIT Press.
- - - . 1983. The Modularity of Mind. Cambridge, Mass.: MIT/Bradford Press.
- - - . 1984. "Information and Association." Unpublished paper, Massachusetts Institute of
Technology.
---;Bever, T. G.; and Garrett, M. 1974. The Psychology of language. New York: McGraw-Hill.
Fodor, J. A.; and Katz, J. J., eds. 1964. The Structure of Language: Readings in the Philosophy of
Language. Englewood Oiffs, N.J.: Prentice-Hall.
Fodor, J. A.; and Pylyshyn, Z. W. 1981. "How Direct Is Visual Perception: Some Reflections
on Gibson's 'Ecological Approach,'" Cognition 9: 139-96.
Foster, J. M. 1967. List Processing. London: Macdonald.
Foucault, M. 1970. The Orrkr of Things. New York: Pantheon.
Fox, J, K. 1983. "Debate on Learning Theory is Shifting," Science 222: 1219-22.
Frazer, J. 1955. The Golden Bough: A Study in Magic and Religion, 3rd ed. New York: St. Martin's
Press. Originally published in 1890.
Freeman, D. 1983. Mr~.rgaret Mead and Samoa: The Mrlldng and Unmaldng of an Anthropological
Myth. Cambridge, Mass.: Harvard University Press.
Fries, C. C. 1963. "The Bloomfield School." In C. Mohrmann, A. Sommerfelt, and J. What-
mough, eds., Trends in Europerm Rnd American Linguistics 1930-1960. Utrecht: Spectrum.
Fritsch, G.; and Hitzig, E. 1870. "Uber die elektrische Erregbarkeit des Grosshims," Arch. AnRt.
Physiol. u. Wisse. Med. 37.
Fryer, D. M.; and Marshall, J. C. 1979. "The Motives of Jacques de Vaucanson," Technology
and Culture 20: 257-69.
Gardin, J. C. 1965. "On a Possible Interpretation of Componential Analysis in Archaeology,"
American Anthropologist 67: 9-11.
Gardner, H. 1975. The Shattered Mind: The Person after Brain Dtlmage. New York: Alfred A. Knopf.
- - . 1978. Developmental Psychology. Boston: Uttle, Brown.
- - - . 1983. FrRmes of Mind: The Theory of Multiple Intelligences. New York: Basic Books.
Gazdar, G. 1981. "On syntactic categories," Philosophical TrRnsactions of the Royal Society of Lon-
don, Series B. 295: 267-83.
Gazzaniga, M. S. 1981. "Nobel Prize for Physiology or Medicine," Science 214: 517-18.
- - - . 1983. "Right Hemisphere Language Following Commisorotomy: A Twenty-Year
Perspective," American Psychologist 38: 525-37.
Geertz, C. 1973. The Interpretation of Cultures. New York: Basic Books.
- - - . 1983. Local Knowledge. New York: Basic Books.
Geschwind, N. 1965. "Disconnexion Syndromes in Animals and Man," Brain 88: 585-644.
- - - . 1974. Selected Papers on Language and the Brain. Dordrecht-Boston/Reidel.
Gibson,]. J, 1950. The Perception of the Vtsual World. Boston: Houghton-Mifflin.
- - - . 1966. The Senses Consirkred as Perceptual Systems. Boston: Houghton-Mifflin.
- - - . 1967. "New Reasons for Realism," Synthese, 17: 162-72.
- - - . 1979. The Ecological ApproRch to Vtsual Perception. Boston: Houghton-Mifflin.
Glass, A. L.; Holyoak, K. J.; and Santa,]. L. 1979. Cognition. Reading, Mass.: Addison-Wesley.
Goldman, P. S.; and Galkin, T. W. 1978. "Prenatal Removal of Frontal Association Cortex
397
References
in the Fetal Rhesus Monkey: Anatomical and Functional Consequences in Postnatal
Life," Brain Research 152: 451-85.
Goldman-Rakic, P. S.; lsseroff, A.; Schwartz, M. L.; and Bugbee, N. M. 1982. "Neurobiology
of Cognitive Development in Non-human Primates." Unpublished manuscript, Yale
School of Medicine, New Haven.
Goldstine, H. 1972. The Computer from PllSCal to oon Neumann. Princeton: Princeton University
Press.
Goodenough, W. H. 1951. Properly, Kin, and Community on Truk. New Haven: Yale University
Press.
· - - - . 1956. "Componential Analysis and the Study of Meaning," Langtmgt 32: 195-216.
- - - . 1964. "Language and Property in Truk: Some Methodological Considerations." In D.
H. Hymes, ed., Language in Culture and Society: A Reader in Linguistics and Anthropology. New
York: Harper & Row.
- - - . 1968. "Componential Analysis." In D. Sills, ed., International Encyclopedia of the Social
Sciences. New York: Macmillan.
Goodman, N. 1955. Fact, Fiction and Forecast. Cambridge, Mass.: Harvard University Press.
- - - . 1972. "The Revision of Philosophy." InN. Goodman, Problems and Projects. Indianapo-
lis: Bobbs-Merrill. Originally published in 1963.
Gross, C. G.; Rocha-Miranda, C. E.; and Bender, D. 1972. "Visual Properties of Neurons in
Inferotemporal Cortex of the Macaque," Journal of Neurophysiology 35: 96-111.
Grossberg, S. 1980. "How Does a Brain Build a Cognitive Code?," Psychological R1!11iew 87: 1-51.
Gruber, H. E.; and Vom!che, J. J., eds. 1977. The Essential Piaget. New York: Basic Books.
Hacking, I. 1980. "Is the End in Sight for Epistemology?," j(}Urnal of Philosophy 77: 579-88.
- - - . 1982. "Wittgenstein the Psychologist" [Review of Remarks on the Philosophy of Psy-
chology, G. E. M. Anscombe, ed.], New York Review of Books, 1 April.
Halacy, D. S., Jr. 1962. Computers the Machines We Think With. New York: Dell.
Hanfling, 0., ed. 1972. Fundamental Problems in Philosophy. Blackwell: Open University Press.
Harris, M. 1968. The Rise of Anthropological Theory: A Htstory of Theories of Culture. New York:
Thomas Y. Crowell.
Harris, R. 1981. The Language Myth. London: Duckworth.
Harris, W. H.; and Levy, J. S., eds. 1975. "Holography." The New Columbia Encyclopedia. New
York: Columbia University Press.
Harris, Z. S. 1952. "Discourse Analysis," Language 28: 18--30.
Harth, E. 1982. Windows on the Mind: Reflections on the Physical Basis of Consciousness. New York:
William Morrow.
Haugeland, J. 1981. "The Nature and Plausibility of Cognitivism." In J. Haugeland, ed., Mind
Design: Philosophy, Psychology, Artificial Intelligence. Montgomery, Vt.: Bradford Books.
Havelock, E. A. 1982. Preface to Plato. Cambridge, Mass.: Belknap Press/Harvard University
Press.
Hayes, J. R. 1978. Cognitive Psychology: Thinking and Creating. Homewood, Ill.: Dorsey Press.
Hayes, P. 1982. Personal communication. March 1982.
Hebb, D. 0. 1949. Organization of Behavior. New York: John Wiley.
Heider, E. R. 1972. "Universals in Color Naming and Memory," ](}U17Ull of frperimenflll Psy-
chology 93: 10-20.
---;and Olivier, D. C. 1972. "The Structure of the Color Space in Naming and Memory
for Two Languages," Cognitive Psychology 3: 337-54.
Heims, S. J. 1980. fohn von Neumann and Norbert Wiener. Cambridge, Mass.: MIT Press.
Helmholtz, H. von. 1962. Treatise on Physiological Optics. Vol. Ill J.P. C. Southall, trans. New York:
Dover.
Herrnstein, R. J.; and Boring, E. G., eds. 1965. A Source Book in the History ofPsychology. Cambridge,
Mass.: Harvard University Press.
Herskovitz, M. J. 1953. Franz Boas: The Science of Man in the Making. London: Charles Scribner.
Hilton, A. M. 1963. Logic, Computing Machines, and Automation. New York: World Publishing.
Hinton, G. E. 1981. "Implementing Semantic Networks in Parallel Hardware." In G. E. Hinton
and J, A. Anderson, eds., Parallel Models of Associative Memory. Hillsdale, N.J.: Lawrence
Erlbaum.
---;and Anderson, J. A. 1981. Parallel Models of Associative Memory. Hillsdale, N.J.: Lawrence
Erlbaum.
Hinton, G. E.; Sejnowski, T. J.; and Ackley, D. H. 1984. "Boltzmann Machines: Constraint
Satisfaction Networks That Learn." Computer Science Memo, Carnegie-Mellon University.
398
References
Hjelmslev, L. 1953. "Prologomena to a Theory of Language," IndiRnR Publimlions in Anthropology
Rnd Linguistics, no. 7.
Hochberg, J. 1978. Perception. Englewood Cliffs, N.J.: Prentice-Hall.
Hockett, C. 1958. A Course in Modem Linguistics. New York: Macmillan.
- - . 196&. Article on Leonard Bloomfield. In D. L. Sills, ed., lnlmzRtionRI Encyc/opediR of tht
SodRI Sdenm, pp. 95-98. New York: Macmillan.
- - . 1968b. Tht Shlte of the Art. The Hague: Mouton.
Hodgen, M. T. 1964. Early Anthropology in the Sixteenth Rnd Sromtemth Centuries. Philadelphia:
University of Pennsylvania Press.
Hodges, A. 1983. Aum Turing: Tht EnigmR. New York: Simon & Schuster.
Hofstadter, D. R. 1983. "Artificial Intelligence: Subcognition as Computation." In F. Machlup
and U. Mansfield, eds., The Study of ln/ormRfion: lnltrdisdplinRry MtsSRgtS. New York: John
Wiley.
--;and Dennett, D. 1981. The Mind's 1 New York: Basic Books.
Holton, G. 1984. "Do Scientists Need a Philosophy?," Tht Times Literary Supplnnml, 2. November
1984, pp. 2.31-34.
Holyoak, K. 1983. "Toward a Unitary Theory of Mind" [Review of J. R. Anderson, The
Architecture of Cognition], Science 2.2.2: 499-500.
Hooper, J. 1982.. Interview with Karl Pribram. Omni, October 1982, pp. 129-35, 169-76.
Horn, B. K. P. 1975. "Obtaining Shape from Shading Information." In P. Winston, ed., The
Psychology of Compultr V'tsion. New York: McGraw-Hill.
Hubel, D. H. 1978. "Vision and the Brain," Bu/lefin of the AmericRn Academy of Arts and
Sdenm 31: 17-28.
- - . 1979.' "The Brain," Scientific American 241: 44-53.
--;and Wiesel, T. N. 1959. "Receptive Fields of Single Neurones in the Cat's Striate
Cortex," fournRl of Physiology 148: 574-91.
Hubel, D. H.; and Wiesel, T. N. 1962. "Receptive Fields, Binocular Interaction and Functional
Architecture in the Cat's Visual Cortex," JournRI of Physiology 160: 10~4.
- - . 1979. "Brain Mechanisms of Vision," Scientific American 241: 150--62.
Hume, D. 1739. A Tmtise of HumRn Nature. London.
- - . 1955. An lnl{uiry Concerning Human Undershlnding. New York: Liberal Arts Press. Original
work published in 1748.
Humphrey, G. 1951. Thinking: An Introduction to ExpmmmtRl Psychology. New York: John Wiley.
Hunt, M. 1982. The Universe Within. New York: Simon & Schuster.
Hutchins, E. 1980. Culture Rnd Inference: A TrobriRnd CRSe Study. Cambridge, Mass.: Harvard
University Press.
Hymes, D. 1974. Studies in the History of Linguistics. Bloomington: Indiana University Press.
lnhelder, B.; and Piaget, J. 1958. Tht Grf!Wth of Logiml Thinking from Childhood to Adolescence, trans.
by A. Parsons and S. Milgram. New York: Basic Books.
- - . 1964. Tht Early Grf!Wth of Logic in the Child: ClllSSi/imtion Rnd SeriRtion. London: Routledge
& Kegan Paul.
Jackendoff, R. 1983. SemRnfics and Cognition. Cambridge, Mass.: MIT Press.
Jackson, J. Hughlings. 1932. Selected Writings. London: Hodder & Stoughton.
Jakobson, R. 1941. KindersprRChe, AphRSie und Rl/gemeine LRutgesetze. Uppsala: Universitets Aarssk-
rift. English ed.: Child LRngugge, AphRSiR Rnd PhonologicRl UnivtrsRls. The Hague: Mouton, 1968.
- - . 1963. Essais de linguislil{ue gblb-Rle. Paris: Editions de minuit.
- - ; and Halle, M. 1956. Furulamenhl/s of LRngu~~ge. The Hague: Mouton.
James, W. 1890. The Principles of Psychology. New York: Henry Holt.
Jefferson, G. 1949 "The Mind of Mechanical Man," Lister oration for 1949, British Mediml
fourn~~l 1: 1105-21.
Jeffress, L. A., ed. 1951. Cerebral Mechanisms in Behar:lior. Tht HIXon Symposium. New York: John
Wiley.
Johnson, M. K.; Bransford, J.D.; and Soloman, S. 1973. "Memory for Tacit Implications of
Sentences," ]ournRI of Ezperimenhll Psychology 98: 203-5.
Johnson-Laird, P. N. 1982. "Propositional Representations, Procedural Semantics, and Mental
Models." In J. Mehler, E.C.T. Walker, and M. Garrett, eds., Perspecfi!lts on MmtRl Represmhl·
lion: Experimmhll Rnd Theoretical Studies of Cognifi'Of Processes and CRpadties. Hillsdale, N.J.: Law-
rence Erlbaum.
- - . 1983. Menhll Models: TOWRrds a Cognifiue Sdmce of LRngu~~gt, Inference, Rnd Consdousness.
Cambridge, Mass.: Harvard University Press.
399
References
---;and Wason, P. C. 1970. "A Theoretical Analysis of Insight into a Reasoning Task,"
Cogniti'Of Psychology 1: 134-48.
Johnson-Laird, P. N.; and Wason, P. C., eds. 1977. Thinking: Readings in Cogniti'Of Science. Cam-
bridge: Cambridge: University Press.
Kahneman, D.; Slovic, P.; and Tversky, A., eds. 1982. judgement under Uncertainty: Heuristics and
Biases. New York: Cambridge University Press.
Kahneman, D.; and Tversky, A. 1982. "The Psychology of Preferences," Scientific American 246:
160-74.
- - - . 1984. "Choices, Values, and Frames," American Psychologist 39: 341-50.
Kandel, E. R. 1979. "Small Systems of Neurons," Scientific American 241: 66--84.
- - - . 1982. "Steps toward a Molecular Grammar for Learning: Explorations into the Nature
of Memory." Paper presented at the Bicentennial Symposium of the Harvard Medical
School, October 1982.
Kant, I. 1958. Critil{lit of Pure Reason, trans. by N. Kemp Smith. New York: Random House.
Original work published in 1781.
Kardiner, A.; and Preble, E. 1961. They Studied Man. Oeveland: World Publishing.
Katz, J.; and Fodor, J. 1963. "The Structure of a Semantic Theory," Language 39: 170-210.
Kay, P. 1966. "Comment on B. N. Colby, Ethnographic Setrulntics: A Preliminary Survey," Current
Anthropo{ggy 7: 3-32.
Keesing, R. M. 1976. Cultural Anthropology: A Contemporary Perspecti'Of. New York: Holt, Rinehart
& Winston.
Keil, F. C. 1979. Semantic and Conceptual Development: An Ontological Perspective. Cambridge, Mass.:
Harvard University Press.
Kenny, A. 1973. Witfgmsfein. Harmondsworth, Middlesex, England: Penguin Books.
Kessen, W. 1981. "Early Settlements in New Cognition," Cognition 10: 167-71.
Kimura, D. 1973. "The Asymmetry of the Human Brain," Scimtific American 228: 70-80.
Kinsbourne, M., ed. 1978. Assymefrical Function of the Brain. Cambridge: Cambridge University
Press.
Kitto, H. D. F. 1951. The Greeks. Middlesex, England: Penguin Books.
Kohler, W. 1925. The MenfRiify of Apes, trans. by E. Winter. New York: Humanities Press.
(Original work published in 1917.)
- - - . 1969. The Task of Gestalt Psychology. Princeton: Princeton University Press.
Konishi, M. 1969. "Experimental Studies in the Ontogeny of Avian Vocalizations." In R. A.
Hinde, ed., Bird Vocalization. Cambridge: Cambridge University Press.
Kosslyn, S.M. 1978. "Imagery and Internal Representation." In E. Rosch and B. B. Lloyd, eds.,
Cognition and Categorization. Hillsdale, N.J.: Lawrence Erlbaum.
- - - . 1980. Image and Mind. Cambridge, Mass.: Harvard University Press.
- - - . 1981. "The Medium and the Message in Mental Imagery: A Theory," Psychological
Review 88: 46-66.
- - - . 1983. Ghosts in the Mind's Machine: Creating and Using Images in the Brain. New York: W.
W. Norton.
- - - . 1984. "Imagery and Cerebral Lateralization: A Computational Theory." Unpublished
paper, Harvard University.
---;Ball, T. M.; and Reiser, B. J, 1978. "Visual Images Preserve Metric Spatial Information:
Evidence from Studies of Image Scanning," journal of Experimental Psychology: Hui1Uln Perception
and Performance 4: 47-60.
Kosslyn, S.M.; Pinker, S.; Smith, G. E.; and Schwartz, S. P. 1979. "On the Demystification
of Mental Imagery," The Behauioral and Brain Scienas 2: 535-81.
Kripke, S. 1972. "Naming and Necessity." In D. Davidson and G. Harman, eds., Stmantics of
Natural Language. Dordrecht/Boston: Reidell.
Kroeber, A. 1917. "The Super-organic," Americ11n Anthropologist 17: 163-213.
- - - . 1948. Anthropology. New York: Harcourt, Brace.
Kroeber, T. 1970. A/fred Kroeber: A Personal Configur11tion. Berkeley: University of California
Press.
Kuhn, T. 1970. The Structure of Scientific Revolution, 2nd ed. Chicago: University of Chicago Press.
Kuper, A. 1973. Anthropologists and Anthropology: The British School 1922-1972. New York: Pica
Press.
Kyburg, H. E. 1983. "Rational Belief," The Behavioral and Br11in Sciences 6: 231-73.
Labbe, R.; Firl, A.; Mufson, E.; and Stein, D. 1983. "Fetal Transplants: Reduction of Cognitive
Deficits in Rats with Frontal Cortex Lesions," Science 221: 470-72.
400
References
Lachman, R.; Lachman, J. L.; and Butterfield, E. C. 1979. Cognitive Psychology and Information
Processing: An Introduction. Hillsdale, N.J.: Lawrence Erlbaum.
Lakoff, G. 1969. "Empiricism without Facts," Foundations of Language 5: 118-27.
- - - . 1980. "Whatever Happened to Deep Structure?" The Behavioral and Brain Sciences 3:
22-23.
- - . 1982. "Categories and Cognitive Models." Working paper. Berkeley Cognitive Sci-
ence Report no. 2.
--;and Ross, J. R. 1976. "Is Deep Structure Necessary?" In J. McCawley, ed., Syntu: and
Semantics, vol. 7. New York: Academic Press. Originally written in 1967.
Lashley, K. S. 1929. Brain Mechanisms and Intelligence. Chicago: University of Chicago Press.
- - . 1950. "In Search of the Engram." In Symposia of the Scdefy for Experimental Biology 4:
454-82.
Lashley, K. S.; Chow, K. L.; and Semmes, J. 1951. "An Examination of the Electrical Field
Theory of Cerebral Integration," Psychological Review 58: 123-36.
Lave, J. 1977. "Cognitive Consequences of Traditional Apprenticeship Training in West
Africa," Anthropology and Education Qu11rferly 8: 177-80.
Leach, E. 1961. "Levi-Strauss in the Garden of Eden: An Examination of Some Recent
Developments in the Analysis of Myth," TransiiCtions of the New York Academy of Sciences
D, 23: 386-96.
- - - . 1974. Uvi-SirRU$5, rev. ed. London: Fontana.
Leary, D. E. 1978. "The Philosophical Development of the Conception of Psychology in
Germany, 1780-1850," JournAl of the Htsfory of the Behavioral Sciences 14: 113-21.
Lees, R. B. 1957. "Review of Noam Chomsky," Syntactic Structures, LRnguage 33: 375-408.
Lenat, D. B. 1976. "AM: An Artificial Intelligence Approach to Discovery in Mathematics as
Heuristic Search." Report STAN-CS-76-570. Stanford University, Computer Science
Department.
- - - . 1984. "Computer Software for Intelligent Systems," Scientific American 251: 204-13.
Lettvin, J. Y. 1981. "Filling Out the Forms: An Appreciation of Hubel and Wiesel," Sci-
ence 214: 518-20.
- - ; Maturana, H. R.; McCulloch, W. S.; and Pitts, W. H. 1959. "What the Frog's Eye Tells
the Frog's Brain," Proceedinss of the IRE 47: 1940-51.
Levi-Strauss, C. 1963. Structural Anthropo/ggy, trans. by C. Jacobson and B. Grundfest Schoepf.
New York: Basic Books. Original work published 1958.
- - - . 1964. Tristes fropiques, trans. by J. Russell. New York: Atheneum. Original work
published in 1955.
- - . 1969. The Raw 11nd the Cooked, trans. by J. and D. Weightman. New York: Harper & Row.
Original work published in 1964.
Ughthill, J. 1972. "A Report on Artificial Intelligence." Unpublished manuscript, Science
Research Council.
- - ; Sutherland, N. S.; Needham, R. M.; Longuet-Higgins, H. C.; and Michie, D. 1973.
Artificial Intelligence: A Paper Symposium. London: Science Research Council.
Locke, J. 1975 An Essay Concerning Human Understanding. P. H. Nidditch, ed. Oxford: Clarendon
Press. Original work published 1690.
Longuet-Higgins, H. C. 1979. "The Perception of Music," Proceedings of the Royal Society of London
series B 205, pp. 307-22.
- - - . 1981. "Artificial Intelligence-A New Theoretical Psychology?," Cognition 10: 197-
200.
- - - . 1982. "A Theory of Vision" [review of David Marr's Vision), Science 218: 991-92.
- - - ; Lyons, J.; and Broadbent, D. 1981. The Psychological Mechanisms of Language. A joint
symposium of the Royal Society and the British Academy, London.
Lorenz, K. 1935. "Der Kumpan in der Umwelt des Vogels," Journal of Ornithology 87: 137-413.
Lounsbury, F. G. 1956. "A Semantic Analysis of the Pawnee Kinship Usage," Language 32:
158-94.
- - - . 1963. "Linguistics and Psychology." In S. Koch, ed., Psychology: A Study of a Science. New
York: McGraw-Hill.
- - . 1968. "One Hundred Years of Anthropological Unguistics." In J. 0. Brew, ed., One
Hundred Years of Anthropology. Cambridge, Mass.: Harvard University Press.
Lovejoy, A. 0. 1964. The Great Chllin of Being. Cambridge, Mass: Harvard University Press.
Lowry, R. 1971. The Evolution of PsychologicRl Theory 1650 to Present. Chicago: Aldine-Ather-
ton.
401
References
Luce, R. D.; Bush, R. R.; and Galanter, N. F., eds. 1963. Handbook of Mathematical Psychol-
ogy. New York: John Wiley.
Luchins, A. S. 1942. "Mechanization in Problem Solving," Psychological Monographs 54(6), whole
no. 248.
Luria, A. R. 1966. Higher Cortical Functions in Man. New York: Basic Books.
Lyons, J. 1968. lnlroduction to Theoretical Linguistics. Cambridge: Cambridge University Press.
- - - . 1970. Noam Chomsky. New York: Viking Press.
McArthur, D. J. 1982. "Computer Vision and Perceptual Psychology," Psychowgical Bulletin 92:
283-309.
McCarthy, J. 1984. Interview, Stanford University, February 1984.
- - ; Abrahams, P. W.; Edwards, D. J.; Hart, T. D.; and Levin, M. I. (1962). Lisp I. 5
Programmn's Manual. Cambridge, Mass.: MIT Press.
---;and Hayes, P. J. 1969. "Some Philosophical Problems from the Standpoint of Artificial
Intelligence." In B. Meltzer and D. Michie, eds., Machine Intelligence 4. Edinburgh: Edin-
burgh University Press.
McCorduck, P. 1979. Machines Who Thinlr. San Francisco: W. H. Freeman.
McCulloch, W.; and Pitts, W. 1943. "A Logical Calculus of the Ideas Immanent in Nervous
Activity," Bulletin of Mathematical Biophysics 5: 115-33.
McDougall, W. 1961. Body and Mind. Boston: Beacon Press. Original publication in 1911.
Malinowski, B. 1961. Argonauts of the Western-Pacific: An Account of Natiue Enterprise and Adventure
in the Archipelagoes of Meltmesian N!!W Guinea. New York: E. P. Dutton. Original work
published in 1922.
- - - . 1968. Culture. In D. L. Sills, ed., International Encyclopedia of the Social Sciences. New York:
Macmillan. Original work published in 1935.
Mandler, G. 1981. "What Is Cognitive Psychology? What Isn't?" Address to the APA Divi-
sion of Philosophical Psychology, Los Angeles.
Mandler, J. M.; and Mandler, G. 1964. Thin/ring: From Association to GesfRlt. New York: John
Wiley.
Marcus, M. 1979. "A Theory of Syntactic Recognition for Natural Language." In P. H.
Winston and R. H. Brown, eds., Artificial Intelligence: An MIT Perspectiue, vol. 1, pp. 191-229.
Cambridge, Mass.: MIT Press.
- - - . 1980. A Theory of Syntactic Recognition for Natural Language. Cambridge, Mass.: MIT Press.
Marie, P. 1906. "Revision de la question de l'aphasie," Semaine MMicale 21: 241-47.
Marler, P. 1982. "Avian and Primate Communication: The Problem of Natural Categories,"
Neuroscience and Biobehavioral Revii!WS 6: 87-94.
- - - ; and Peters, S. 1977. "Selective Vocal Learning in a Sparrow," Science 198: 519-21.
Marr, D. 1969. "A Theory of Cerebellar Cortex," journal of Physiology 202: 437-470 (London).
- - - . 1979. "Representing and Computing Visual Information." In P. H. Winston and R.
H. Brown, eds., Artificial Intelligence: An MIT Perspectiue, vol. II: Understanding V"tsion, Manipula-
tion, Computer Design, Symbol Manipulation. Cambridge, Mass.: MIT Press.
- - - . 1982. V"tsion: A Computational Investigation info the Human Representation and Processing of Visual
Information. San Francisco: W. H. Freeman.
---;and Nishihara, K. 1978. "Representation and Recognition of the Spatial Organization
of Three-Dimensional Shapes," Proceedings of the Royal Society London, B, 200, pp. 269-94.
Marshall, J. C. 1980. "Artificial Intelligence-The Real Thing?," The Behavioral and Brain
Sciences 3: 435-37.
- - - . 1981. "Cognition at the Crossroads" [review of M. Piattelli-Palmarini, ed., Language
and Learning: The Debate between Jean Piaget and Noam Chomsky], Nature 289: 613-14.
Matthews, J. R. 1982. "Who's Afraid of Immanuel Kant?" [review of N. Block, ed., Readings
in Philosophy of Psychology, vols. I, II], Contemporary Psychology 27: 433-34.
Mehler, J.; Morton, J.; and Jusczyk, P. 1984. "On Reducing Language to Biology," Cognitiue
Neuropsychology 1: 83-116.
Mehler, J.; Walker, E. C. T.; and Garrett, M., eds. 1982. Perspectives on Mental Representation:
Experimental and Theoretical Studies of Cognitiue Processes and Capacities. Hillsdale, N.J.: Lawrence
Erlbaum.
Mervis, C. B.; and Rosch, E. 1981. "Categorization of Natural Objects," Annual Review of
Psychology 32: 89-115.
Miller, G. A. 1956. "The Magical Number Seven, Plus or Minus Two: Some Limits on Our
Capacity for Processing Information," Psychological Revii!W 63: 81-97.
- - - . 1962. "Some Psychological Studies of Grammar," American Psychologist 17: 748-62.
402
References
- - - . 1979. "A Very Personal History." Talk to Cognitive Science Workshop, Massachu-
setts Institute of Technology, Cambridge, Mass. 1 June 1979.
- - - . 1982. Interview, Princeton University, January 1982.
- - - ; Galanter, E.; and Pribram, K. 1960. Plans and the Structure of Behavior. New York: Holt,
Rinehart & Winston.
Miller, J. 1983. Slates of Mind. New York: Pantheon Books.
Milner, B. 1967. "Brain Mechanisms Suggested by Studies of Temporal Lobes." In F. L.
Darley, ed., Brain Mechanisms Underlying Speech and Language. New York: Grune & Strat-
ton.
Milner, P. M. 1970. Physiological Psychology. New York: Holt, Rinehart & Winston.
Minsky, M. 1963. "Steps toward Artificial Intelligence." In E. A. Feigenbaum and J. Feldman,
eds., Computers and Thought. New York: McGraw-Hill. Originally circulated in 1957.
- - - , ed. 1968. Semantic Information Processing. Cambridge, Mass: MIT Press.
- - - . 1975. "A Framework for Representing Knowledge." In P. H. Winston, ed., The
Psychology of Computer Vision. New York: McGraw-Hill.
- - . 1979. "The Society Theory." In P. H. Winston and R. H. Brown, eds., Artificial
lnfelligence: An MIT Perspective, vol. 1. (pp. 423-50). Cambridge, Mass.: MIT Press.
- - - . 1982. "Learning Meaning." Unpublished manuscript, Massachusetts Institute of
Technology.
---;and Papert, S. 1968. Percepfrons. Cambridge, Mass.: MIT Press.
Mishkin, M. 1967. "Visual Mechanisms beyond the Striate Cortex." In R. W. Russell, ed.,
Frontiers in Physiological Psychology. New York: Academic Press.
Moates, D. R.; and Schumacher, G. M. 1980. An Introduction to Cognitive Psychology. Belmont,
Calif.: Wadsworth.
Mohrmann, C.; et al. 1963. Trmds in Modern Linguistics. Utrecht: Spectrum.
Moray, N. 1969. Lisfming and Attention. Baltimore, Md.: Penguin.
Morris, C. 1971. Writings on the Gmeral Theory of Signs. The Hague: Mouton.
Morton, J. 1981. "Will Cognition Survive?" Cognition 10: 227-34.
Mountcastle, V. B. 1978. "An Organizing Principle for Cerebral Function: The Unit Module
and the Distributed System." In G. M. Edelman and V. B. Mountcastle, eds., The Mindful
Brain. Cambridge, Mass.; MIT Press.
Murray, H. A. 1945. "Assessment of the Whole Person," Proceedings Meeting Military Psychologists
and Psychiatrists. University of Maryland Press.
Murray, S. 0. 1982. "The Dissolution of 'Classical Ethnoscience,' "fourmzl of fhe History of fhe
Behavioral Sciences 18: 63-175.
Neisser, U. 1967. Cognitive Psychology. New York: Appleton-Century-Crofts.
- - - . 1976. Cognition and Reality. San Francisco: W. H. Freeman.
- - - . 1980. "Comments on Gestalt Psychology." Paper presented at APA Symposium,
September 1980.
- - - . 1984. "Toward an Ecologically Oriented Cognitive Science." Emory Cognition Project,
Report 1, Atlanta, Georgia, January 1984.
Newell, A. 1983. "Intellectual Issues in the History of Artificial Intelligence." In F. Machlup
and U. Mansfield, eds., The Study of Information: Interdisciplinary Messages. New York: John
Wiley.
- - - ; Shaw, J. C.; and Simon, H. A. 1964. "Elements of a Theory of Human Problem-
solving." In R. J. C. Harper, C. C. Anderson, C. M. Christensen, and S.M. Hunka, eds.,
The Cognitive Processes: Readings. Englewood Cliffs, N.J.: Prentice-Hall. Reprinted from Psy-
chological Review 65 (1958): 151-66.
Newell, A.; and Simon, H. A. 1972. Human Problem Solving. Englewood Cliffs, N.J.: Prentice-
Hall.
Newmeyer, F. 1980. Linguistic Theory in America: The First Quarter Century of Transformational Genera-
five Grammar. New York: Academic Press.
Nilsson, N. J. 1980. Principles of Artificial Intelligence. Palo Alto, Calif.: Tioga.
Norman, D. 1980. "Twelve Issues for Cognitive Science,'' Cognitive Scimce 4: 1-32.
Nottebohm, F. 1970. "Ontogeny of Bird Song," Science 167: 950-56.
- - - . 1980. "Brain Pathways for Vocal Learning in Birds: A Review of the First 10 Years,''
Progress in Psychobiological and Physiological Psychology 9: 85-124.
Osherson, D. N.; and Smith, E. E. 1981. "On the Adequacy of Prototype Theory as a Theory
of Concepts," Cognition 9: 35-58.
Paivio, A. 1971. Imagery and Verbal Processes. New York: Holt, Rinehart & Winston.
403
References
Peters, S.; and Ritchie, R. 1973. "On the Generative Power of Transformational Grammars,"
Information Sciences 6: 49-83.
Piaget, J. 1970. "Piaget's Theory." In P. Mussen, ed., Carmichael's Manual of Child Psychology
l New York: John Wiley.
Piattelli-Palmarini, M., ed. 1980. Language and laming. Cambridge, Mass.: Harvard University
Press.
- - . 1983. "The Parliament of Mind" [review of J. A. Fodor, The Modularity of Mind], The
Times Lilerflry Suppltmmf, 2 December 1983, p. 1357.
Pinker, S. 1984. Language Uflmability and languflgt DnM/opment. Cambridge, Mass.: Harvard Uni-
versity Press.
Plato. 1956fl. Mmo. In E. H. Warmington and P. 0. Rouse, eds., Great Dialogues of Plato, trans.
by W.H.D. Rouse. New York: New American Library.
- - . 1956b. Thfletdus. In I. Edman, ed., The Philosophy of Plato, trans. by B. Jowett. New York:
Random House.
Popper, K. 1974. Unendd Quest: An Intellectual AutobWgmphy. Glasgow, Scotland: William Col-
lins.
Posner, M. 1969. "Abstraction and the Process of Recognition." In G. H. Bower and J. T.
Spencer, eds., The Psychology of laming and Moti'(}(lfion, vol. 3. New York: Academic Press.
- - - ; and Mitchell, R. F. 1967. "Chronometric Analysis of Classification," Psychological
Rtview 74: 392-409.
Posner, M.; and Shulman, G. L. 1979. "Cognitive Science." In E. Hearst, ed., The First Century
of Experimenftll Psychology, Hillsdale, N.J.: Lawrence Erlbaum.
Pribram, K. H. 1971. LangullgtS of the BrRin: Experimenftll Paradoxes Rnd Principles in Neuropsy-
chology. Englewood Cliffs, N.J.: Prentice-Hall.
- - - . 1980. "Mind, Brain, and Consciousness: The Organization of Competence and Con-
duct." In R. J. Davidson and M. Davidson, eds., The Psychology of Consciousness. New York:
Plenum Press.
- - - . 1982. "Localization and Distribution of Function in the Brain." In J. Orbach, ed.,
Neuropsychology after Lashley. Hillsdale, N.J.: Lawrence Erlbaum.
Putnam, H. 1960. "Minds and Machines." InS. Hook, ed., Dimensions of Mind. New York: New
York University Press.
- - - . 1973. "Philosophy and Our Mental Life." Paper presented at Foerster Symposium on
Computers and the Mind at University of California at Berkeley. In H. Putnam, Mind,
Language and Reality: Philosophical Papers, vol. 2. Cambridge: Cambridge University Press,
1975.
- - . 1975a. Mind, Language 11nd Reality: Philosophical PRpers, vol. 2. Cambridge: Cambridge
University Press.
- - . 1975b. "The Meaning of 'Meaning."' (Reprinted in Putnam, 1975, pp. 215-71.)
- - - . 1981. "Why Reason Can't Be Naturalized." Howison Lecture, University of Califor-
nia at Berkeley, 30 April1981. Reprinted in Putnam, 1983.
- - - . 1983. Realism and Reason: Philosophical P11pers, vol. 3. Cambridge: Cambridge University
Press.
- - - . 1984. "After Ayer, after Empiricism," PRrtisan Review 2: 265-75.
Pylyshyn, Z. W. 1973. "What the Mind's Eye Tells the Mind's Brain: A Critique of Mental
Imagery," Psychological Bulletin 8: 1-14.
- - - . 1979. "Imagery Theory: Not Mysterious-Just Wrong," The BehlluiOrRl 11nd Brain
Sciences 2: 561-62.
- - - . 1980a. "The 'Causal Power' of Machines" [commentary on J. Searle, "Minds, Brains
and Programs'1, The &h11vioraland Br11in Sciences 3: 442-44.
- - - . 1980b. "Cognitive Representation and the Process-architecture Distinction," The
Beh11uioral Rnd BrRin Sciences 3: 154-69.
- - - . 1981. "The Imagery Debate: Analogue Media versus Tacit Knowledge," Psychological
Review 88: 16-45.
- - - . 1984. Compuftltion and Cognition: ToWRrd a Foundation for Cognitiut Science. Cambridge, Mass.:
MIT Press.
Quillian, M. R. 1968. "Semantic Memory." In M. Minsky, ed., Semantic Information Process-
ing. Cambridge, Mass.: MIT Press.
Quine, W. V. 0. 1953. From a Logical Point of View. Cambridge, Mass.: Harvard University Press.
Radcliffe-Brown, A. R. 1940. "On Social Structure," founu~l of the Royal Anthropologic11l Insti-
tute 70.
404
References
- - - . 1952. Structure Rnd Function in PrimifiTJe Society. Glencoe, n.: Free Press.
Rapaport, D. 1951. OrgtmizRfion Rnd PRthology of Tlwught. New York: Columbia University Press.
Reddy, D. R.; Erman, L. D.; Fennell, R. D.; and Neely, R. B. 1973. "The HEARSAY Speech
Understanding System." In Proceedings, Third InternRfion~tlfoint Conference on ArfijiciRl Intelli-
gence 3: 185-93 (Stanford).
Rey, G. 1981. "What Are Mental Images?" InN. Block, ed., lm11gery. Cambridge, Mass.: MIT
Press.
Rieber, R. W. 1983. DiRlogues on the Psychology of LRngullge Rnd Thought: ConversRtions with NORm
Chomsky, ChRrles Osgood, /l!lln PiRget, Ulric Ntisser, Rnd MRrcel Kinsbourne. New York: Plenum
Press.
Rivers, W. H. R. 1900. "A Genealogical Method of Collecting Social and Vital Statistics,"
fournRl of the AnthropologicRl lnstilute of Grl!llt BritRin Rnd Ireumd 30: 74-82.
Roberts, L. G. 1965. "Machine Perception of Three Dimensional Solids." InT. T. Tippett et
al., OpticRl Rnd Electro-opficRl lnformRtion Processing, Cambridge, Mass.: MIT Press.
Robins, R. H. 1967. A Short History of Linguistics. Bloomington: Indiana University Press.
Robinson, D. M. 1976. An lntelleciuRl History of Psychology. New York: Macmillan.
Romney, A. K.; and D' Andrade, R. G. 1964. "Cognitive Aspects of English Kin Terms,"
AmericRn Anthropologist 66: 146-70.
Rorty, R. 1979. Philosophy Rnd the Mirror of NR!ure. Princeton, N.J.: Princeton University Press.
- - - . 1982R. Personal communication. Princeton University, January 1982.
- - - . 1982b. "Mind as Ineffable." Unpublished manuscript, Princeton University.
Rosch, E. 1973R. "Natural Categories," CognitiTJe Psychology 4: 328-50.
- - - . 1973b. "On the Intemal Structure of Perceptual and Semantic Categories." InT. E.
Moore, ed., CognitirJe Development Rnd the A(fuisifion of LRngullgt. New York: Academic Press.
- - - . 1977. "Human Categorization." In N. Warren, ed., Aduances in Cross-culiuml Psychol-
ogy, vol. I. London: Academic Press.
- - - . 1978. "Principles of Categorization." In E. Rosch and B. B. Lloyd, eds., Cognition Rnd
CRiegorizRtion. Hillsdale, N.J.: Lawrence Erlbaum.
--;and Lloyd, B. B. 1978. Cognition Rnd GltegorizRtion. Hillsdale, N.J.: Lawrence Erlbaum.
Rose, S. 1973. The Conscious Bmin. New York: Alfred A. Knopf.
Rosenblueth, A.; Wiener, N.; and Bigelow, J. 1943. "Behavior, Teleology, and Purpose,"
Philosophy of Science 10: 18-24.
Rosenfield, I. 1984. "Seeing through the Brain," New York ReuinD of Books, 11 October 1984, pp.
53--56.
Royce, J. R.; and Rozeboom, W. W., eds. 1972. The Psychology of Knowing. New York: Gordon
& Breach.
Rumelhart, D. E. 1975. "Notes on a Schema for Stories." In D. G. Bobrow and A. M. Collins,
eds., RepresentRtion Rnd UnderstRnding. New York: Academic Press.
- - - . 1982. "Comments on Cognitive Science." Unpublished manuscript, University of
California at San Diego.
- - - ; and McClelland, J. L. 1982. "An Interactive Activation Model of Context Effects in
Letter Perception: part 2. The Contextual Enhancement Effect and Some Tests and
Extensions of the Model," PsychologicRl R1!11iew 89: 60-94.
Russell, B. 1961. History of Western Philosophy. London: George Allen & Unwin.
Ryle, G. 1949. The Concept of Mind. London: Hutchinson.
Sahlins, M. 1976. "Colors and Cultures," SmriotiCR 16: 1-22.
Sapir, E. 1921. lRngullgt. New York: Harcourt, Brace, & World.
- - - . 1929. "The Status of Linguistics as a Science." In Selected Writings of EdWRrd SRpir.
Berkeley: University of California Press.
Saussure, F. de. 1959. Course in Geneml Linguistics, ed. by C. Bailey and A. Sechehaye, in
collaboration with A. Riedlinger; trans. by W. Baskin. New York: Philosophical Library.
Originally published in 1915.
Schank, R. C. 1972. "Conceptual Dependency: A Theory of Natural Language Understand-
ing," CognitiTJe Psychology 3: 552-631.
- - - . 1980. "An Artificial Intelligence Perspective on Chomsky's View of Language," The
BehR~iOrRl Rnd Bmin Scimces 3: 35-37.
---;and Abelson, R. 1977. Scripts, PIRns, GORis, Rnd Undershlnding. Hillsdale, N.J.: Lawrence
Erlbaum.
Schwartz, R. 1981. "Imagery-There's More to It Than Meets the Eye." InN. Block, ed.,
f=gery. Cambridge, Mass.: MIT Press.
405
References
Schwartz, S. P. 1979. "Natural Kind Terms," Cognitiqn 7: 301-15.
Scribner, S. 1984. "Studying Working Intelligence." In B. Rogoff and J. Lave, eds., Ewryday
Cognitiqn. Cambridge, Mass.: Harvard University Press.
--;and Cole, M. 1981. The Psychology of Literacy. Cambridge, Mass.: Harvard University
Press.
Searle, J. 1980. "Minds, Brains and Programs," The Beha~ioral and Brain Sciences 3: 417-57.
- - - . 1983a. lnlentionalily: An Essay in the Philosophy of Mind. Cambridge: Cambridge University
Press.
- - - . 1983b. "Minds and Brains without Programs." Unpublished essay.
Segall, M. H.; Campbell, D. T.; and Herskovits, M. J. 1966. The Influence of Culture on VJSual
Perceptillrl. Indianapolis: Bobbs-Merrill.
Shannon, C. E. 1938. "A Symbolic Analysis of Relay and Switching Circuits." Master's thesis,
Massachusetts Institute of Technology; published in Trans11ctiqns of lhe American lnstilufe of
Electrical Engineers 57: 1-11.
--;and McCarthy,}., eds. 1956. AutomafR Studies; Annals of Mathematical Studies # 34. Prince-
ton: Princeton University Press.
Shanon, B. 1985. "The Role of Representation in Cognition." In J. Bishop, J. Lockhead, D.N.
Perkins, eds., Thinking. Hillsdale, N.J.: Lawrence Erlbaum.
Shebar, W. 1979. "Mental Imagery: A Critique of Cognitive Psychology." Paper submitted
for A.B. with Honors in Philosophy, Harvard University, Cambridge, Mass.
Shepard, R.N. 1981. "Psychophysical Complementarity." In M. Kubovy and J. R. Pomerantz,
eds., Perceptual Organizatillrl. Hillsdale, N.J.: Lawrence Erlbaum.
- - . 1982. "Perceptual and Analogical Bases of Cognition." In J. Mehler, E. C. T. Walker,
and M. Garrett, eds., Perspecti~ on MmfRl RepresenfRtiqn: EzperimenfRI11nd Theoretical Studies of
Cogniti~ Processes and Cap~~cities. Hillsdale, N.J.: Lawrence Erlbaum.
- - - ; and Chipman, S. 1970. "Second-order Isomorphism of Internal Representations:
Shapes of States," Cogniti~ Psychology 1: 1-17.
Shepard, R. N.; and Metzler, J. 1971. "Mental Rotation of Three-dimensional Objects,"
Science 171: 701-3.
Shiffrin, R. M.; and Schneider, W. 1977. "Controlled and Automatic Human Information
Processing: II. Perceptual Learning, Automatic Attending, and a General Theory," Psycho-
logiclll Review 84: 127-90.
Silverstein, M. 1978. "The Three Faces of Function: Preliminaries to a Psychology of Lan-
guage." In M. Hickmann, ed., Proceedings of a Working Conference on the Social Foundations of
Lllnguagt 11nd Thought. Chicago: Center for Psychosocial Studies.
Simon, H. A. 1969. Sciences of the Ar/ifici11l. Cambridge, Mass.: MIT Press.
- - - . 1982. Interview, Carnegie-Mellon University, February 1982.
---;and Newell, A. 1958. "Simulation of Cognitive Processes: A Report on the Summer
Research Training Institute, 1958." In Social Science Research Council Items 12: 37-40.
Skinner, B. F. 1957. Verbal Behllvior. New York: Appleton-Century-Crofts.
Sloan Foundation. 1976. "Proposed Particular Program in Cognitive Sciences." New York.
- - - . 1978. "Cognitive Science, 1978." Report of the State of the Art Committee. New
York.
Sloman, A. 1978. The Computer Rroolution in Philosophy: Philosophy, Science and Models of Mind.
Hassocks, Sussex: Harvester Press.
Smith, E. E.; and Medin, D. L. 1981. Categories and Concepts. Cambridge, Mass.: Harvard Univer-
sity Press.
Smith, N.; and Wilson, D. 1979. Modtrn Linguistics: The Results of Chomsky sRroolutillrl. Middlesex:
Penguin Books.
Sperber, D. 1968. Que'esl-ce que le structur11lismel 3. Lt Strucluralismt en anthropologie. Paris: Seuil.
- - - . 1982. Lt sawir des anlhropologues. Paris: Hermann.
Sperling, G. 1960. "The Information Available in Brief Visual Presentations," Psychological
Monographs 24: whole no. 11.
Sperry, R. W. 1951. "Mechanisms of Neural Maturation." InS. S. Stevens, ed., Handbook of
Ezperimenl11l Psychology. New York: John Wiley.
- - - . 1974. "Lateral Specialization in the Surgically Separated Hemispheres." In F. Schmitt
and F. G. Worden, eds., The Neurosciences: Third Study Program. Cambridge, Mass.: MIT Press.
- - - ; and Miner, N. 1955. "Pattern Perception Following Insertion of Mica Plates into
Visual Cortex," founu~l of Comparatiw 11nd Physiological Psychowgy 48: 463-69.
Sternberg, S. 1966. "High-Speed Scanning in Human Memory," Science 153: 652-54.
406
References
- - - . 1969. "Memory-Scanning: Mental Processes Revealed by Reaction-Time Experi-
ments," American Sdenfist 57: 421-57.
Stich, S. P. 1982. "The compleat cognitivist" [review of J. A. Fodor, Rqmsenftltions: Philosophical
Essays on the Foundations of Cognitiue Science], Confempora'71 Psychologist 27: 419-21.
- - . 1983. From Folk Psychology irJ Cognitiue Sdmce. Cambridge, Mass.: MIT Press (Bradford
Books).
- - - . 1984. "Thinking as per program" [review of P. N. Johnson-Laird, Mental Models:
Towrmis a Cognitive Sdence of Language, Inference, and ConsciOIISness ], Times Lifert~ry Supplement, 4
February 1984, p. 189.
Stocking, G. W. 1982. Rau, Culture, and Eoolufion: Essays in the Hislrlry of Anthropology. Chicago:
University of Chicago Press.
Stouffer, S. A.; et al. 1949. The American Soldier. Princeton, N.J.: Princeton University Press.
Sussman, G. J. 1975. A Computer Model of Skill Acl{uisifion. New York: American Elsevier.
Sutherland, S. 1982. "The Vision of David Marr" [review of D. Marr, VISion], Nature 298:
691-92.
Thomason, R., ed. 1974. Formal Philosophy: Selected Papers of Richard Montague. New Haven: Yale
University Press.
Thomson, R. 1968. The Pelican Htstory of Psychology. Middlesex, England: Penguin.
Tinbergen, N. 1951. The Study of lnslinct, London: Oxford University Press.
Tolman, E. C. 1932. Purposiue Behauior in Animals and Men. New York: Century.
Treisman, A.M. 1960. "Contextual Cues in Selective Listening," Quarterly Journal of Ezperimenflll
Psychology 12: 242-48.
- - - . 1964. "Selective Attention in Man," British Medical Bulletin 20: 12-16.
Turing, A.M. 1936. "On Computable Numbers, with an Application to the Entscheidungs-
Problem," Proceedings of the London Mathematical Somty, Series 2, 42: 230-65.
- - - . 1963. "Computing Machinery and Intelligence." In E. A. Feigenbaum and J. Feldman,
eds., Computers and Thought. New York: McGraw-Hill. Original work published 1950.
Turvey, M. T.; Shaw, R. E.; Reed, E. S.; and Mace, W. M. 1981. "Ecological Laws of Perceiving
and Acting: In Reply to Fodor and Pylyshyn (1981)," Cognition 9: 237-304.
Tversky, A.; and Kahneman, D. 1983. "Extensional vs. Intuitive Reasoning: The Conjunction
Fallacy in Probability Judgment," Psychological Review 90: 293-315.
Tyler, S. A., ed. 1969. Cognitiue Anthropology. New York: Holt, Rinehart, & Winston.
- - - . 1978. T1te Said and the Unsaid: Mind, Meaning, and Culture. New York: Academic Press.
Tylor, E. 1871. Primitive Culture. London: John Murray.
Ullman, S. 1980. "Against Direct Perception," T1te Behauioral and Brain Sdmces 3: 373-416.
Vachek, J. 1966. The Linguistic School of Prague. Bloomington: Indiana University Press.
Van Lehn, K.; Seely Brown, J.; and Greeno, J. 1982. Competitive Argumentation in Compuflllionttl
Theories of Cognition (Interim report). Cognitive and Instructional Science Series. Xerox
Pare, Palo Alto.
Von Neumann, ]. 1958. The Computer and the Brain. New Haven, Conn.: Yale University
Press.
Waldrop, M. 1984a. "Artificial Intelligence (1): Into the World," Science 223: 802-5.
- - - . 1984b. "The Intelligence of Organizations," Science 225: 1136-37.
- - . 1984c. "Artificial Intelligence in Parallel," Science 225: 608-10.
Wallace, A. F. C. 1968. "Cognitive Theory." InD. L. Sills, ed., lnfernllfional Encyclopedia of the Social
Sciences. New York: Macmillan.
--;and Atkins, J. 1960. "The Meaning of Kinship Terms," American Anthropologist 62:
58-80.
Waltz, D. 1975. "Understanding Line Drawings of Scenes with Shadows." In P. H. Winston,
ed., The Psychology of Computer VISion. New York: McGraw-Hill.
Wason, P. C. 1966. "Reasoning." In B. Foss, ed., New Horizons in Psychology, vol. 1. Harmonds-
worth, Middlesex: Penguin.
- - ; and Johnson-Laird, P. N. 1972. The Psychology ofReasoning: Structure in Content. Cambridge,
Mass.: Harvard University Press.
Watson, J. B.; and Rayner, R. 1920. "Conditioned Emotional Reactions," Journal of Experimental
Psychology 3: 1-14.
Watson, R.I., ed. 1979. Basic Writings in the Hisirlry of Psychology. New York: Oxford University
Press.
Weimer, W. B. 1973. "Psycholinguistics and Plato's Paradoxes of the Meno," American
Psychologist, January, pp. 15-33.
407
References
Weiss, P. 1952.. "Central versus Peripheral Factors in the Development of Coordination,"
Research Publications of the Association for Research in Nm10us and Mental DistaSe 30: 3-2.3.
Weizenbaum, J. 1966. "ELIZA-A Computer Program for the Study of Natural Language
Communication between Man and Machine," Communications of the Association for Computing
Machinny 9: 36-45.
- - - . 1976. Computer Power and Human Rtason. San Francisco: W. H. Freeman.
Wernicke, C. 1874. Der aphasische sympfomen!ompla. Breslau: Cohn & Weigart.
Werthheimer, M. 1912.. "Experimentelle Studien iiber das Sehen von Bewegungen," Utshrift
for Psychologie 61: 161-2.65.
- - - . 1945. Productive ThinJ:ing. New York: Harper & Brothers.
Wexler, K. 1978. "A review of John Anderson, Liznguage, MetnllrY and Thought," Cognition 6:
32.7-51.
- - - . 1982.. "A Principle Theory for Language Acquisition." In L. Gleitman and H. E.
Wanner, eds., Liznguagt AclfUisition: The Stale of the Art. New York: Cambridge University
. Press.
--;and Culicover, H. 1980. Formal Principles of Language Acquisitwns. Cambridge, Mass.: MIT
Press.
White, L.A. 1963. "The Ethnography and Ethnology of Franz Boas," Memorial Museum Bulletin,
no. 6, Austin, TX.
Whitehead, A. N. 192.5. Science and the Modern World. New York: Macmillan.
- - - ; and Russell, B. 1910-13. Principia Mathnnatica. Cambridge: Cambridge University
Press.
WhorE, B. 1956. Liznguage, Thought and Reality, ed. by}. Carroll. Cambridge, Mass.: MIT Press.
Wiener, N. 1961. Cybernetics, or Control and Communication in the Animal and the Machine, Znd ed.
Cambridge, Mass.: MIT Press. Original work published in 1948.
- - . 1964. Ex-prodigy. Cambridge, Mass.: MIT Press.
Wilensky, R. 1983. Planning and Understanding: A Computational Approach to Human Rta5oning.
Reading, Mass.: Addison-Wesley.
Wilson, M. D., ed. 1969. The Essential Descartes. New York: Mentor.
Winograd, T. 1972.. Understanding Natural Language. New York: Academic Press.
- - - . 1975. "Frame Representation and the Declarative-Procedural Controversy." In D. G.
Bobrow and A. Collins, eds., Representation and Undnstanding: Studies in Cognitive Science. New
York: Academic Press.
Winston, P. H. 1977. Artifidallntelligenct. Reading, Mass.: Addison-Wesley.
---;and Brown, R. H., eds. 1979. Artificial Intelligence: An MIT Perspective, vol. 2.: Underslanding
Vision, Manipulation, Computer Design, Symbol Manipulation. Cambridge, Mass.: MIT Press.
Witherspoon, G. J. 1971. "Navajo Categories of Objects at Rest," American Anlhropologist 73:
110-17.
Wittgenstein, L. 1961. Tradalus logico-philosophicus, trans. by D. F. Pears and B. F. McGuinness.
London: Routledge & Kegan Paul. Original work published in 192.1.
- - - . 1968. Philosophical Investigations, trans. by G. E. M. Anscombe. New York: Macmillan.
Original work published in 1953.
Wolff, R. P., ed. 1967. Kant. New York: Doubleday.
Woods, B. T. 1980. "Observations on the Neurological Basis for Initial Language." In D.
Caplan, ed., Biological Studies of Mental Processes. Cambridge, Mass.: MIT Press.
Woods, W. A. 1975. "What's in a Link: Foundations for Semantic Networks." In D. G.
Bobrow and A. Collins, eds., Representation and Underslanding: Studies in Cognitive Science. New
York: Academic Press.
Young, R. M. 1974. "Production Systems as Models of Cognitive Development." In Proceedings
of AISB Summer Conference. University of Sussex.
Zaidel, E. 1977. "Lexical Organization in the Right Hemisphere." In P. Busser and A. Rougel-
Buser, eds., Cerebral Correlales of Conscious Experience. Amsterdam: Elsevier.
---.1978. "Auditory Language Comprehension in the Right Hemisphere Following Cere-
bral Commissurotomy and Hemispherectomy: A Comparison with Child Language and
Aphasia." In A. Caramazza and E. B. Zurif, eds., Language Acquisition and Liznguagt Breakdown:
Parallels and Dim-gencits. Baltimore: Johns Hopkins University Press.
Zangwill, 0. L. 1980. "Kenneth Craik: The Man and His Work" (The Kenneth Craik Lecture,
10 March 1978), British journal of Psychology 71: 1-16.
408
NAME INDEX
409
Name Index
Chomsky (confinutd} Fodor, Jerry, riv, 34, 65, 81-86, 132, 215, 242,
75, 81, 84, 86, 125, 140, 168, 170, 182-96, 312-13, 314, 315, 332, 387, 391
199-200, 202, 205, 207-28, 239, 241, 242, Foucault, Michel, 291
253,264,293,306,312,314,332,382,386, Franz, Shepherd Ivory, 260--61
391 Frazer, Sir James, 229-30
Cohen, Gillian, xiu Frege, Gottlob, 16, 23, 60, 61, 227, 292, 360
Cohen, L. Jonathan, 373--78 Freud, Sigmund, 15, 75, 166, 196, 380, 389
Colby, Kenneth, 156-57, 158, 159, 161 Fries, Charles, 204
Cole, Michael, riv, 254, 256 Fries, Jakob Friedrich, 99
Conklin, Harold, 30 Fritsch, Gustav Theodor, 267, 284
Copernicus, Nicholas, 59
Craik, Kenneth, 16, 293
Galanter, Eugene, 32
Galileo, 75
D'Andrade, Roy,riu, 34,248, 25~57, 369-70 Gall, Francis Joseph, 266, 267, 278, 284
Darwin, Charles, 8, 14, 38, 227, 228, 291, 392 Galvani, Luigi, 266
Dejerine, Jules, 267--68 Gardner, Hilde Weilheimer, u
Democritus, 296 Gardner, John, 32
Dennett, Daniel, riu, 79-81, 82, 160, 178 Gardner, Ralph, v
Descartes, Rene, 4, 43, 49-57, 66--67, 72, 73, Gazdar, Gerald, 217-18, 220
74,75,81,82,85,86,87,142,192,196,216, Gazzaniga, Michael, riu
224, 238, 265, 278, 284, 291, 323, 392 Geertz, Clifford, riv, 243--44, 250, 253, 258,
Dewey, John, 73, 77, 109 355-56, 358, 359, 380, 388
Dodwell, Peter, 74, 85, 86 Gerard, Ralph, 23
Donders, F. C., 101, 102, 121, 132 Geschwind, Norman, riu, 275
Dresher, Elan, 177 Gibson, James J., 118, 294, 296, 307, 308-16,
Dreyfus, Hubert, riv, 81, 163, 164, 387 319, 320, 321, 322, 332, 346, 358
Duncker, Karl, 113 Gilson, Etienne, 49
Durant, Will, 56 Glass, Arnold, 134
Gleitman, Henry, 347
Gleitman, Lila, 347
Glucksberg, Samuel, riv
Goldstein, Kurt, 268
Goodenough, Ward, 30, 245, 248
Ebbinghaus, Hermann, 104-5, 115, 116, 126- Goodman, Nelson, riu, 64, 65, 351
27 Goodnow, Jacqueline, 29, 93
Eccles, Isabel, xu Gregory, Richard L., 269
Edgerton, Robert B., 246 Greissman, Judith, xu
Ehrenfels, Christoph von, 111 Grimm, Jakob, 197, 198
Einstein, Albert, 60, 113 Grodzinsky, Josef, riu
Euclid, 65, 139 Gross, Charles, riu
Evans, T. G., 151-52
410
Name Index
Heider, Eleanor, see Rosch, Eleanor Keesing, Roger, 252
Helmholtz, Hermann von, 99-102, 117, 292, Kessen, William, 130
296, 305, 307, 308 Kessler, Martin, ru
Herbart, Johann Friedrich, 99 Keyser, Samuel Jay, riu
Herodotus, 226 Klivington, Kenneth, riii
Herskovits, Melville, 254 Kliiver, Heinrich, 23, 272
Hinton, Geoffrey, riu, 167, 387 Koffka, Kurt, 111
Hitzig, Eduard, 267, 284 Kohler, Wolfgang, 111, 112-14, 271, 272-73
Hjelmslev, Louis, 196 Kortz, Nan, rv
Hockett, Charles F., 204, 206, 208 Kosslyn, Stephen, riv, 129, 294, 325, 326-36,
Hofstadter, Douglas, 176 330-32, 366, 382
Holyoak, Keith, 131, 134 Krechevsky, Mara, rv
Hornstein, Norbert, 177 Kripke, Saul, 351-52, 353, 360
Hoss, Phoebe, ru Kroeber, Alfred, 234
Huarte, Juan, 265 Kuhn, Thomas, 209
Hubel, David, 30, 273-74, 282, 283, 284-85, Kiilpe, Oswald, 105, 106
302, 304, 307 Kyburg, Henry, 378
Hull, Clark, 110
Hume, David, 53, 54, 55-56, 57, 61, 81, 87,
291
Hunt, Morton, 36
Hutchins, Edwin, 256
Lakoff, George, riv, 209, 212
Langness, L. L., 246
Lashley, Karl, 11, 12-14, 15, 20, 110, 118,
186, 191, 192-93, 260-64, 268, 270, 272,
273, 276, 278, 283, 382
lnhelder, Barbel, 377 Lave, Jean, riu, 256
Isard, Stephen, riu Leach, Edmund, 241
lsay, Jane, ru Leary, David, 99
Lederberg, Joshua, 155
Lees, Robert, 186, 189-90
Leibnitz, Gottfried, 57
Lenat, Douglas, 171
Lenneberg, Eric, 343, 344
Jackson, Hughlings, 269 Lettvin, Jerome, riu, 30, 274, 307
Jakobson, Roman, 26, 189, 196, 200--2, 205, Levine, Linda, riv
219, 221, 235-36, 245 LeVine, Robert, riv
James, William, 107-9, 292 Levi-Strauss, Claude, riu, 26, 75, 140, 195,
Johnson, Samuel, ru 202, 235, 236-44, 257-59, 293, 350
Johnson-Laird, Philip, riu, 294, 331-32, 361- Levy-Bruhl, Lucien, 223-25, 257-59
70, 373, 379, 380, 382 Lighthill, Sir James, 164
Jones, Sir William, 196-97 Locke, John, 4, 53, 54-55, 57, 61, 71, 72, 73,
Jung, Carl, 380 77, 83, 87, 105, 291
Loffredo, Carmella, ru
Longuet-Higgins, Christopher, riu, 171, 179-
80, 306
Lorenz, Konrad, 31
Lounsbury, Floyd, 30, 245, 248
Kagan, Jerome, riv Lowie, Robert, 228
Kahneman, Daniel, 361, 370-79, 400 Luchins, Abraham, 113
Kandel, Eric, 279-80, 281-82 Luria, Alexander, 16, 26, 276
Kant, Immanuel, 4, 53, 56-60, 61, 65, 68, 71,
72,73,75,76,86,87,98-99, 100,114,116,
117, 118, 137, 192, 291, 392
Kaplan, Ronald, riv
Kardiner, Abram, 229
Katz, Jerrold, 34, 215 McCarthy, John, riu, 23, 30, 138-39, 145,
Kay, Paul, riu, 252-53, 255, 294, 348-49, 154-55, 178
350, 358 McClelland, Jay, riv
411
Name Index
McConnell, Susan, .riv Norman, Donald, .riu, 36, 387
McCorduck, Pamela, 138 Nottebohm, Fernando, 2.80-82.
McCulloch, Warren, 10,18-19, 2.0, 2.1, 2.2., 2.4,
2.6, 30, 139, 144, 148, 2.72.-73, 307, 32.0
McDougall, William, 118, 2.30
Magendie, Fran.;ois, 2.66
Malcolm, Norman, 74 Olivier, Donald, 344
Malinowski, Bronislaw, 2.33-34, 2.92. Oppenheimer, }. Robert, 2.4, 2.5, 95
Mandler, Jean, riv, 2.9 Osherson, Daniel, riu, 347-48, 350
Marbe, Karl, 106 Osthoff, H., 197
Marcus, Mitchell, 169
Marie, Pierre, 2.68
Marler, Peter, 2.81
Marr, David, 170-71, 2.16, 2.94, 2.98-308, 310,
311, 319, 32.0, 32.1-2.2., 382., 384 Paivio, Allan, 330
Marshack, Alexander, riu Panini, 196
Marshall, John, xiv, 142., 179, 180, 2.82. Papert, Seymour, 169
Marx, Karl, 75 Parisi, Domenico, .riu
Maturana, Humberto, 30 Paul, Hermann, 197
Mehler, Jacques, xiu Pavlov, Ivan, 12.
Messer, August, 106 Peano, Cduseppe, 60
Metrodorus, 2.96 Peirce, Charles Sanders, 60
Mettrie, Julian Offray de Ia, 142. Peters, Stanley, riu
Metzler, Jacqueline, 12.9, 32.5 Piaget, Jean, 2.6, 75, 83,116-18,119,132.,136,
Meyer, Christine, .riv 140, 2.14, 2.19, 2.42., 2.54, 2.64, 2.92., 377,
Meyers, C. S., 2.30 389
Millar, Susanna, .riu Pinker, Steven, 2.15, 330
Miller, George, .riv, 2.3, 2.4, 2.6, 2.8-2.9, 32., 64, Pitts, Walter, 18-19, 2.0, 2.1, 2.6, 144, 32.0
89-91, 92., 95, 97, 110, 116, 119, 132., 140, Planck, Max, 60
195, 2.14, 2.39, 2.64, 2.93, 382. Plato, 3, 4, 52., 76, 77, 86, 191, 192., 2.12., 2.91,
Minsky, Marvin, .riv, 2.5, 2.6, 30, 34, 139, 140, 2.96, 358, 392.
145, 151-54, 161, 165-66, 167, 169 Popper, Karl, 76-77
Mishkin, Mortimer, 2.74 Posner, Michael, .riv, 2.9, 32.
Montague, Richard, 64, 2.16 Preble, Edward, 2.2.9
Montesquieu, Charles Louis, 2.2.6-2.7, 2.2.8, Pribram, Karl, .riv, 32., 2.64, 2.70, 2.83
2.41 Putnam, Hilary, .riv, 31, 66, 77-78, 81, 82., 88,
Moravcsik, Julius, riv 195, 351-52., 353, 358, 388
Moray, Neville, 12.0 Pylyshyn, Zenon, 175, 2.94, 312.-13, 314, 315,
Morgan, Lewis Henry, 2.45 332.-36, 368, 387, 391
Morris, Charles, 2.45
Morton, John, riu, 134, 2.82.
Moses, Joel, 161
Mountcastle, Vernon, 2.74
Muller, Johannes, 2.66 Quillian, Ross, 169
Murray, Stephen, 2.52.-53 Quine, Willard V. 0., 65, 66, 70-71, 74, 195,
351
412
Name Index
Rosenblatt, Frank, 307 Talmy, Leonard, .riu
Ross, J. R., 209 Tarski, Alfred, 65
Rousseau, Jean-Jacques, 241 Thorndike, Edward L., 12, 110
Rumelhart, David, .riv, 125 Tinbergen, Niko, 31
Russell, Bertrand, 17, 23, 60, 61, 62, 64, 65, Titchener, Edward Bradford, 105, 107,
73, 143, 145-46, 291, 360 108
Ryle, Gilbert, 66--68, 69, 70, 71, 80, 81, Tolman, Edward C., 110, 118
82 Treisman, Anne, 120
Troubetskoy, Nikolay, 200
Turing, Alan, 16, 17-18, 25, 26, 139, 144, 156,
188, 293, 382, 385
Turvey, Michael, 314, 315, 316
Sahlins, Marshall, 349-50, 356 Tversky, Amos, 294, 360, 361, 370-79, 389
Samuel, Arthur, 139 Tyler, Stephen, 34, 245, 251
Santa, John, 134 Tylor, Edward, 227-29, 231, 256, 292
Sapir, Edward, 183, 195, 204, 205, 234-35
Saussure, Ferdinand de, 196, 198-200, 202,
219, 221, 232, 236, 292
Schank, Roger, riu, 23, 165, 167-68, 177-78,
216
Scheffler, Israel, riv Ullman, Shiman, 311-12, 314, 321
Schlick, Morris, 63
Schwartz, Stephen, 352
Scribner, Sylvia, 256
Searle, John, .riu, 81, 140, 168, 171, 172-76,
177, 199, 249
Segall, Marshall, 254 van Lehn, Kurt, 178
Shannon, Claude, 21, 22, 144, 188, 382 Vaucanson, Jacques de, 142
Shanon, Benny, xi~ 123,388 Vico, Giambattista, 4
Shaw, Cliff, 146, 147 Voegelin, C. F., 190
Shaw, George Bernard, 229 von Ehrenfels, ste Ehrenfels, Christoph von
Shaw, Robert, 314, 315, 316 von Helmholtz, Ste Helmholtz, Hermann von
Shebar, William, 336-38 von Humboldt, Wilhelm, 196
Shepard, Roger, 128--30, 132, 324-25, 330, von Neumann, John, 10, 17, 18, 19, 21, 22, 23,
338 24,26,29,139,145,180,188,307,319,320,
Shiffrin, Richard, 122-23 382, 385
Shulman, Gordon, 32 Vygotsky, Lev, 26
Siegler, Robert, riv
Silverstein, Michael, 250
Simon, Herbert, .riu, 22, 23, 25, 26, 28, 29, 30,
33, 34, 64, 139, 140, 145-51, 154, 155, 160,
161, 179, 219, 241, 264, 293, 319, 320, 356, Wagner, Sheldon, riu
361, 382, 385, 386 Wallace, Anthony, 246-49, 250
Simon, Theodore, 116 Waltz, David, 170, 297
Skinner, B. F., 12, 80, 110, 191-93, 194, 310, Wanner, Eric, riii
312, 314 Wason, Peter, 361-62, 369, 373, 379
Sloman, Aaron, .riu, 138 Watson, John B., 12, 109-11, 260, 264, 292
Slovic, P., 373 Watt, Henry, 106
Smith, Brian Cantwell, .riu Weaver, Warren, 21
Smith, Edward, 347-48, 350 Weiss, A. P., 203
Socrates, 3, 4 Weiss, Paul, 23, 270
Spence, Kenneth, 110 Weizenbaum, Joseph, 156-57, 158, 162, 163,
Sperber, Daniel, 242-43, 380 177
Sperling, George, 120 Wernicke, Carl, 267, 284
Sperry, Roger, 270, 275-76 Wertheimer, Max, 111-12, 113, 114
Stephens-Swannie, Laura, xu Wexler, Kenneth, 177
Sternberg, Saul, 121-23 Whately, Bishop, 227
Strauss, Claudia, riu White, Leslie, 233
Sussman, Gerald, 160 Whitehead, Alfred North, 3, 17, 23, 52-53,
Sutherland, Stuart, riv, 307 60, 61, 62, 64, 143, 145-46, 291, 360
413
Name Index
Whorf, Benjamin, 205, 2.32., 2.35, 2.55, 344, Wolf, Connie, ru
357,358 Woods, William, 169
Wiener, Norbert, 15, 2.0, 2.1, 2.2., 2.3, 2.4, 2.6, Wundt, Wilhelm, 102.--4, 105-7, 108, 117,
115, 139, 144, 188, 2.93, 2.95, 307 136, 198, 203, 2.2.7, 2.92., 32.4
Wiesel, Torsten, 30, 2.73-74, 2.82., 2.83, 2.84-
85, 302., 304, 307
Wilensky, Robert, 141
Winner, Ellen, .riu
Winograd, Terry, Iiu, 158-60, 161, 162., 164,
165, 177, 178, 2.16, 2.97 Young, Richard, 171
Winston, Patrick, 170
Witherspoon, Gary, 2.50
Wittgenstein, Ludwig, 62., 63, 68-70, 71, 73,
74, 76, 77, 2.86-87, 336-39, 342., 346, 351,
353, 357, 358 Zurif, Edgar, .riu
414
SUBJECT INDEX
ablation: Karl Lashley's use of, 261; tech- 78; Dartmouth College meeting, 30, 138-
nique, 261, 273; of the visual cortex, 261 40; definition of, 140; development of,
Adaptive Control of Thought (ACf), 131- 138-39; and intentional systems, 80; and
32; process of, 131 knowledge, 161; and language, 167-69,
adequacy, Noam Chomsky's standards of, 216-17; and MPPS, 319; and perception,
189 169-71, 304, 319; and problem solving,
adhesion, 228 160-61; and psychology, 135,137, 180-81;
affordances, 310, 312, 315, 317 and reasoning, 379; SHRDLU, 158-60;
agnosia, 22 tensions in, 141; top-down approach, 165-
Alfred P. Sloan Foundation, riii, 35-38, 220; 66; see also computers, information process-
SOAP, 36-37 ing
algorithmic level (David Marr), 299-300 Asptcls of tht Thtory of Synftlx (Chomsky),
American Psychological Association, 97 207-8; responses to, 209
anthropology,26,223-59,341,350,354,356, assemblers, 145
379; in America, 213; case studies, 225; and avunculate, 236-37, 240
classification, 341, 355-56, 358; cognitive,
see cognitive anthropology; and cognitive
science, 41, 44; and color terms, 255; and
computers, 243-44; definition of culture,
228; ethnoscience, 225; functionalism, 227, baseline (or base) rates, 374
233-34; and illusion studies, 254; and lan- basic level, 346
guage,234-36,255;andlearning,228;and Bayes theorem, and the law of statistics,
linguistics, 202, 222, 232, 234-36; and 374-76
mathematical skills, 256; methodology in, behaviorism, 10--16; 23, 26, 65, 86, 94, 109-
228; myths, 239-41; and primitive cul- 11, 119, 191-92, 193, 214, 281, 282, 292-
tures, 223-25, 230, 238-39; and primitive 93, 323,324, 381-82; in America, 113, 292-
mentality, 223, 230, 232; and psychology, 93; and classification, 94; and cognitive
230, 254; and religion, 229; scientific, 231; science, 381; and computers, 10--11; and
symbolic, 355-56; and symbolic processes, the cybernetic approach, 32; differences in
243-44; and transformationalism, 242; and societies, 226; and ethology, 31; and Ge-
visual perception, 254; see also componen- stalt psychology, 113; and introspection,
tial analysis, ethnoscience, ethnosemantic 11-12, 109; in linguistics, 191-92, 203-4,
approach, kinship, structural anthropol- 214; and localization, 270; and logical em-
ogy piricism, 67; in mental imagery, 323; and
Antipodeans, 74, 75, 78 neuroscience, 261, 262, 265; in psychol-
aphasia, 22; and localization, 267 ogy, 109-11, 130, 135, 260; and represen-
Aplysia mli/ornica, 279-80, 282, 287 tational level, 39; revolution, 109; of a sys-
apperception, 103 tem, 332; see also reductionism
artificial intelligence, 5, 40-41, 42, 44, 54, 80, Bible, 226
87, 130, 131, 138-81, 194, 216, 217, 298, bifurcation, 248
304-5,306,379, 388; beginnings of, 142; in bird songs, 280--82
Britain, 164; and cognitive science, 40; and bit (binary digit), 21
computers, 41; critics of, 81, 162-65, 177- bottom-up approach: in Gestalt psychology,
415
Subject Index
bottom-up approach (continued} cognitive anthropology, 30, 244, 245, 250,
112; and David Marr, 305; in psychology, 252; and language, 250-51
97 cognitively penetrable, 333
brain, 31, 264, 275, 332, 384; and forms of Cognifivt Psychology (Neisser), 33, 133
behavior, 284; as a holographic process, cognitive psychology, 42, 44, 95-98, 119-20,
283-84; and perception, 112,300,306, 307; 130; and Adaptive Control of Thought
see also hemispheres, mind (ACT), 131; and artificial intelligence, 180-
brain injury, 269, 277-78, 331; and the im- 81; and bottom-up approach, 97, 126; fu-
age-generation component, 331; and local- ture of, 134; and Gestalt psychology, 114-
ization, 267; and the nervous system, 284 15; history of, 119; and information
Brain Mechanisms and lnfelligmce (Lashley), 261 processing,95-96,119-21,130-34;andin-
formation theory, 96; and memory, 122,
131 (set also memory); and mental repre-
sentations, 128; methodology of, 97-98;
and modular analysis, 132-33; molecular
calculator, 142 vs. molar, 97; problems in, 96; top-down
California Institute of Technology, 10 approach, 97, 124-28; set also psychology
Cambridge Anthropological Expedition, 230, cognitive science: critics of, 42; definition of,
253 6; representations, 38; symptoms, 38
Carnegie Corporation, 32 color: componential analysis, 251; lexicons,
case studies, 225 349
categories of thought, 58 color naming, 324, 343, 344-50; and classical
categorization, 340-59; and anthropology, view of categories, 342; in the Dani, 344;
341, 355-56, 358; boundaries of, 346; and and nervous system, 348
the classical view, 341-42, 345, 347-48; color terms, 255, 258; basic, 350
and cognitive science, 358-59; and color compilers, 145
lexicons, 349; and color naming, 342-50; componential analysis, 244, 246-51, 253,
critics of classical view, 345, 347-48; and 257-58; and kinship, 246-49
empiricism, 342; and ethnoscience, 358; computation, 16-18
extensions of, 342; intensions of, 342; and computational paradox, 9, 44, 133, 180,
language, 351, 354, 358; and logic, 351; 384-88
natural view of, 341, 347, 348; and the ner- computatiopal theory, 299
vous system, 341, 348; and perception, Computer anatht Brain, Tht (von Neumann), 29
349-50, 354, 358, 359; and philosophy, computer languages, 369; LISP, 154-55, 161;
340, 341-42, 350-55; prototypes, 346; and machine, 146
psychology, 344-48; and scientists, 357; computer programs, 138, 166, 256, 329, 334,
structure of, 346-47; and symbolic anthro- 366-67; DENDRAL, 155-56, 161; ELIZA,
pology, 355-56; and Whorf-Sapir hypoth- 157; T. G. Evans, 151-52; GPS, 34, 148-51,
esis, 343 154; HACKER, 160; Logic Theorist (LT),
cells: complex, 273; hypercomplex, 273; sim- 145-48; STUDENT, 152-53, 157; Turing
ple, 273 machine, 18; on visual analysis, 152, 299
Center for Advanced Study in the Behavioral computers, 5, 12, 15, 29, 40-41, 87, 95, 118-
Sciences (Stanford University), 25 24, 134, 145-51, 171, 172, 300, 328-29,
Center for Cognitive Studies (Harvard Uni- 333, 354, 361, 366-67, 379-80, 381, 384-
versity), 32 88; analog, 20; assemblers, 188; and
Center for the Study of Language and Infor- behaviorism, 10-11; in Britain, 16; and
mation (Stanford University), 217, 218 cognitive science, 40-41, 381, 382, 384-88;
centralists, 133 and data structures, 328-29; difference
Cerebral Mechanisms in Behavior Confer- machines, 142; digital, 44, 130, 167, 180,
ence, set Hixon Symposium 264, 320, 325, 334, 385; electronic, 6, 10,
children: and learnability theory, 215; Jean 21, 321; and functionalism, 78; and inten-
Piaget's study of, 116-17, 377; and syllo- tional systems, 79-81; Johnniac, 146; list
gisms, 366 processing, 146; and logic, 361, 366-67;
Chinese Room problem (John Searle), 172- and memory, 30, 34,119, 121-22, 146; and
74, 249; critiques of, 173-74 mental imagery, 328-29, 332, 334; and
classification, 238-39, 292; and behaviorism, mental representation, 328-29; and the
94; classical features of, 341-42; and em- model of the mind, 6, 10, 40, 41, 44, 82,
piricism, 56; and psychology, 93-95 148, 166, 264, 320, 331, 332, 335, 354, 380,
Cognition and Reality (Neisser), 134 388; and perception, 297-98, 300, 307; and
Cognition Project (Harvard University), 93 psychology, 118-24; and representations,
416
Subject Index
3~7; simulation of, 2.45, 32.8-30; space 61--62.; and mental representations, 82.-83;
allocation for, 146; symbolic form of. 34, and neurophysiology, 75-76; in philoso-
12.9, 149; and John von Neumann, 2.9, 145, phy, 49, 53, 60, 71-81; and psychology,
180, 319, 32.0, 385; set also artificial intelli- 74-75; and rationalism, 81; see also knowl-
gence, information processing, informa- edge, philosophy
tion theory equipotentiality, 2.62., 2.70
computer science, 91, 134, 2.72. Eskimos, Franz Boas's fieldwork, 2.31
concept formation, 12.6 Establishment (in perception), 313-18; and
conceptual systems, 2.55 Jerry Fodor, 313-15; Gibsonians' attack of,
connotations, 2.50 316-18; and Zenon Pylyshyn, 313-15; rep-
Critique of Pure Reason (Kant), 43, 57, 59, 392. resentations, 313; and Robert Shaw, 314-
culture, 2.56-57, 343; and color naming, 343, 15, 316; terminology, 315, 317; and Mi-
349; and language, 2.35-36; linear evolu- chael Turvey, 314-15, 316
tion of, 2.31; Edward Tylor's definition of, ethnography, 2.52.
2.2.8 ethnoscience, 2.2.5-2.6, 2.44-53, 355; and cate-
Cybernetics (Wiener), 2.0 gories, 358; and cognitive anthropology,
cybernetics, 19-2.1, 2.4, 32., 144-45, 2.43, 2.45; 2.44, 2.45; and componential analysis, 2.44,
conference on, 2.4; and Norbert Wiener, 2.46-49, 2.51; decline in, 2.52.-53; and eth-
2.0, 144-45 nosemantics, 2.44; goals of, 2.46; and kin-
ship, 2.46-49; and linguistics, 2.46; research
in, 2.52.
ethnosemantic approach, 30, 2.44, 2.46, 2.52.-
53, 2.58
Dani, color studies, 344 ethology, 31
Dartmouth College Meeting, 30, 138-40, Euclid's principles, 65, 139
145-58 evolution, 2.31, 2.32., 2.33; linear, 2.31
declarative representations, 161--62. external properties, 332.
deep structure, 2.07-8, 2.09, 2.12.
DENDRAL, 155-56, 161
determining tendency, 106
difference machine, 142.
direct perception, 311, 312. field theory, 2.71
Discourst on Method, The (Descartes), 392. finite-state grammar, 185, 186, 193
0-structure, 2.18 Ford Foundation, 2.5
foundational philosophy, 56-59; and "cate-
gories of thought," 58; and Immanuel
Kant, 56-59; and knowledge, 57-59; and
the mind, 57; and psychology, 59; and rep-
ecological approach, 315, 316, 317-18; and resentation, 59; and schemas, 58-59; and
the Establishment, 314-18; of James }. the sensory world, 58; and synthetic a priori,
Gibson, 314-15; of Ulric Neisser, 318 57
effective procedure, 367 functional architecture, 333
effectivities, 315, 317 functionalism, 107-9, 162.
ELIZA, 157, 162. functionalism (philosophical), 53, 78-79,
empiricism, 8, 53-.56, 57, 62--63, 81-82., 86, 251, 2.92.; in anthropology, 2.2.7, 2.33-34;
88, 192., 195, 2.14-16, 2.91, 317, 361; agen- and computers, 78; and epistemology, 82.;
da, 56; a priori, 57; British, 342.; and classifi- and the mind-body problem, 78-79
cation, 56, 342.; and introspection, 54, 58, fuzzy set logic, 347
60; and language, 56; and molar analysis,
97; and perception, 55, 309; and philoso-
phy, 52., 54, 353; and psychology, 71, 101;
and Vienna Circle, 65; see also logical em-
piricism General Problem Solver (GPS), 34, 148-51,
engram, 2.62--63 154, 179, 32.0
epistemology, 19, 60, 61--62., 70, 71-81, 81- generative grammar, 187, 194, 2.10, 2.16
86, 87, 116-18, 382.; agenda, 62.; and artifi- Gestalt psychology, 111-18, 119, 12.6, 12.8,
cial intelligence, 31; critics of, 86-87; and 12.9,2.62.,2.65,2.70,271,2.76,2.92.,32.1;anal-
cultural relativism, 77; and empiricism, 81; ysis of, 112.; and artificial intelligence, 113;
and functionalism, 82.; genetic, 117-18; is- and behaviorism, 113; and holism, 112.,
sues in, 14; and language, 82.-84; and logic, 114, 2.69, 2.70-71; and learning, 113; and
417
Subject Index
Gestalt psychology (confinutd) inference, 100; and philosophy, 62
localization, 271; and logical formalisms, information processing: 51, 90-91, 119-20,
117-18; and memory, 115-16; and move- 121,123,125,129,132,134,148,193,239,
ment, 111-12; and perception, 111-12, 305, 307, 311, 318; and architecture, 167;
114,199, 296; and schemas, 116; and struc- Atkinson-Shiffrin model, 122-23, 124-25;
turalism, 113 in Britain, 91-93; and the classical model,
Golden Bough, The (Frazer), 229 128; and cognitive psychology, 95-96, 134;
grammar, 206, 221; finite-state, 185, 186, and computers, 325; criticism on approach,
193; generative, 187, 194, 210, 216; lexi- 133; devices, 132; early systems, 119-21;
cal-functional, 217; phrase-structure, 186; functional architecture, 333; and language,
story, 125; transformational, 186-88, 215, 124-25; limitations, 90; and memory, 121-
216, 218; universal (U.G.), 211-12 24; and mental representations, 128-30;
Greeks, 3-5, 7, 14, 50, 53, 75, 87, 324, 340, and modular analysis, 132-33; and psy-
356, 363, 381, 385; on behavior, 265; and chology, 90-91, 119-21, 122-24, 130-32,
cognitive science, 42.-43; on definition of 311; systems, 40; top-down perspective,
man, 360; epistemology, 14; on grammar, 124-30; see also artificial intelligence, com-
196; on knowledge, 4-5, 291, 296; on per- puters
ception, 296; philosophers, 294; on ratio- information theory, 21-22, 119, 128, 243,
nality, 294 245; and Colin Cherry, 91; and Oaude
Shannon, 21; and Norbert Wiener, 21
Institute for Advanced Study (Princeton
University), 24
intelligence: in apes, 113; in computers, 154;
and learning, 113
HACKER, 160 intentionalism, and philosophy, 53
Harvard University: Center for Cognitive intentional systems, 79-81; and artificial in-
Studies, 32; Cognition Project, 93; philoso- telligence, 80; and computers, 79; Daniel
phers at, 65; Society of Fellows, 25-26 Dennett on, 79-80; Zenon Pylyshyn on,
HEARSAY, 169 175-76; and John Searle, 175-76; Turing
hemispheres, 267-68, 275, 276-78, 284; in machine, 79
birds, 281; effects of injuries on, 277; func- internal representations, 337-38
tions of, 275; and holism, 275; and the im- International Business Machines (IBM), 139
age-generation component, 331; and lan- intrinsic properties, 332
guage, 276-78; left, 275, 276-78, 281, 331; introspection: and behaviorism, 11-12; and
right, 275, 276-78, 281; see also neuro- cognitive psychology, 101, 119; and em-
science piricism, 54; and Hixon Symposium, 11-
Hixon Symposium, 10-16, 18, 23, 29, 110, 12; in philosophy, 50; and Edward Brad-
119, 264, 278, 293, 361, 382; and behavio- ford Titchener, 107; and Wilhelm Wundt,
rism, 11-12; importance of, 25; interdisci- 102-3, 107
plinary studies of, 42; and introspection, IQ test, 116; and Jean Piaget, 116
11-12; and neuroscience, 272-73
holism, 268-70, 271, 275-76, 278, 279, 284,
292; and Gestalt psychology, 268-69; and
Richard Gregory, 269; and hemispheres,
275; and Hughlings Jackson, 269; vs. local-
ization, 270; and nervous system, 268-70; Josiah P. Macy Foundation, 23-24, 36; Sym-
and psychology, 268-69 posium, 42, 293
holography, 283-84 junggrammatilcer, see neo-grammarians
human behavior, 171
human body (mechanical), 51
418
Subject Index
science, 6; and empmc1sm, 54; and the America, 202; and anthropology, 202,232,
Greeks, 4-5, 291; origins of, 54; and phi- 234-36; and artificial intelligence, 216; and
losophy, 43, 52; representational systems, autonomy, 208; and behaviorism, 191-92,
162; study of, 298; see also epistemology 203-4; and cognitive science, 220; and col-
or naming, 344; diachronic, 198; history of,
197; and information processing, 193; and
leamability, 215, 217; and neo-grammari-
ans, 197-98; and philology, 198; and phi-
Language (Bloomfield), 204-5 losophy, 215-16; and phonemes, 26,
language, 14, 27, 62, 68, 69-70, 8.3--84, 104, 200-1, 236; and phonology, 200-2; and
168--69, 182-96, 209, 213, 217-18, 220, phylogeny, 197; and the Prague School,
221,222,235,251,287,292,306,314,336, 200-2; and procedures, 206; and psy-
339, 345, 354; and anthropology, 222, 232, cholinguistics, 214-15; and psychology,
234-36, 237-38, 250-51; and artificial in- 192, 193, 201, 214, 219-20; and rational-
telligence, 167-69, 177,216, 217; and brain ism, 379; and right hemisphere, 275; and
damage, 277-78; and categorization, 351, Standard Theory, 207; structuralism, 206,
354, 358; and cognitive anthropology, 208, 236; theory of, 194; and universal
250-51; comparative, 198; culture, 235-36; grammar, 210; see also Chomsky, language
debates in, 209; and deep structure, 207-8, LISP, 154-55, 161
209, 212; and empiricism, 55; families, 187; list processing, 146
and finite-state grammar, 185, 188, 193; as localization, 261, 262, 263, 265-72, 272-73,
a set of games, 68; and generative gram- 273-74, 275, 278, 284-85, 292; and behav-
mar, 187; and generative semanticists, 209; ior, 270; and the Hixon Symposium, 272;
in hemispheres, 267-68, 275-76, 277-78; vs. holism, 270, and language, 267-68; in
and interpretive semanticists, 209; and nervous system, 284-85; and plasticity,
kernel sentences, 187; learning of, 193, 284; and reductionism, 268; in visual sys-
278, 312; and localization, 267-68; and tem, 270
mental imagery, 336, 339; and neuro- logic, 16, 60, 61, 225, 331, 351, 360, 364-65,
science, 267-68, 276-78, 287; and nota- 367-70; capabilities of, 362; and computer
tions, 203; and philosophy, 69-70, 212; programs, 366-67; and computers, 361; de-
phonemes in, 27, 200-1, 236; and phrase- ductive processes of, 361; in history of
structure grammar, 186, 187, 188; and philosophy, 360; laws of formal, 365; ma-
primitive mentality, 232; and representa- chines, 363; and mental representations,
tional level, 39; and structuralists, 191, 368--69; and model of neuron, 361; and
195; and surface structures, 208, 209, 211; neural networks, 18; pre-logic, 223-24;
and transformational analysis, 186-87; and the primitive mind, 225; and psychol-
and transformational grammar, 187, 188; ogy, 361, 363-64, 370; and reasoning, 363,
and transformations, 187, 206-7; see also 369; in solving problems, 370-72; and syl-
linguistics, semantics, sentences, syntax logisms, 363-64, 365-66, 367-70; see also
Language of Thought, The (Fodor), 83 rationality, reasoning
langue, 196, 199 logical empiricism, 62-66, 67-71, 86-87, 292,
Latin, 196-97 351; criticism of, 66, 86-87; and language,
law of maximal contrast, 201 62-64, 65, 68, 69; and logical syntax, 64;
law of participation, 224 and the mind, 67; and physicalism, 63; and
law of probability, 372, 373, 374, 375 psychology, 68; verificationism, 63; and
learning, 132, 261-63, 271-72, 279-80, 287; the Vienna Circle, 63-65; see also empiri-
and anthropology, 228; in Aplysill californica, cism, philosophy
279-80, 282; and Gestalt psychology, 113; Logic Theorist (LT), 145-48
of language, 193; and language recovery, logic theory machine, 28
276; limitations, 90; and linguistic record-
ing, 90; and natural behavior, 282; after
surgical lesions, 261; theory, 215, 217, 282
Leipzig School, 107
lesion, 261, 268; technique of, 261
lexical-functional theory, 217 Macy Foundation, see Josiah P. Macy Foun-
lexical semantics, 220 dation
lexicons, 211, 212 "Magic Number Seven, Plus or Minus Two,
L 'Homme Machine (Mettrie), 142 The" (Miller), 29, 89
Lincoln Laboratories, 125 markedness, 202
linguistics, 44, 54, 90, 182-222, 232, 379; in Marxism, 233
419
Subject Index
massive parallel processing system (MPPS), "Minds, Brain, and Programs" (Searle),
319-21 171
mathematics, 17, 19, 50, 60-61, 300; and an- Mirror of Nature, set Philosophy and the Mirror
thropology, 256; and cognitive science, 16- of Nature
18; and logic, 61; Markov processes, 26 modal model, 122
melody, 111 modularists, 133
memory, 95, 115-16, 121-24, 125-28, 279; molar analysis, 100, 102, 116, 135; definition
and ACT, 131; Atkinson-Shiffrin model, of, 96-97; and empiricists, 97; and Gestalt
122; and brain damage, 262; and cognitive psychologists, 113
psychology, 122, 131; declarative, 131; and molecular analysis, 100, 101, 109, 135; defini-
functional architecture, 333; and Gestalt tion of, 96-97; and Gestalt psychologists,
psychology, 115-16; holographic theory 112; and rationalists, 97
of, 284; and information processing, 121- Moral Sciences Club, 76
24, 126; long-term, 122; and mental imag- movement: perception, 111-12; and Wert-
ery, 328--29; a model of, 121; and MPPS, heimer, 111-12
319; musical, 131; production, 131, 132; myths, 239-41; autochthony, 240; Oedipus,
and psychology, 121-24, 126-28, 131; 239, 240
schemas, 116; short-term, 28, 122-23;
stores, 122; working, 131; see also comput-
ers: memory
Mm11 (Plato), 3, 4, 43, 72, 83, 291, 392
mental imagery, 50, 323-39, 358; and be-
haviorism, 323; as cognitively penetrable, nativism, 194
333; and computers, 328--29, 332, 334; crit- natural-kind terms, 352-53
ics of Stephen Kosslyn's model of, 330-31; neo-grammarians, 197-98; and psychology,
and data structures, 328--29; flexibility, 198
326; and functional architecture, 333; and nerve-cell functioning, 279
the image generation component, 331; in- nervous system, 13, 19, 20, 260-63, 266, 267,
terest in, 324; and internal representations, 269,273-74,276,278-79,28Z-83;ablation
337-38; and Kosslyn's model and brain of, 261; and behavior, 10; and categories,
damage, 331; and language, 336, 339; and 341; and cells, 273; cognitive capacities of,
logic, 331-32; and map-location task, 326, 22, 41; and color naming, 348; and devel-
333; and memory, 328; and mental models, opment, 273-74; historical views of, 265-
332; and mental representations, 327, 330; 66; holism, 268--70, 271; and holography,
and nerve cells, 338; and neuroscience, 28Z-84; and lesion technique, 261; localiz-
330; and perception, 331; and philosophy, ers vs. holists, 271-72, 278-79; and logical
323-24, 329, 330, 336; and psychology, calculus, 320; and mental processes, 40;
130, 323-25, 327, 328, 329, 331, 336; and model of, 263; and neural commitment,
Zenon Pylyshyn's attack of Stephen Kos- 278; and neural functioning, 265-67; and
slyn, 332-34; and quasi-pictorial, 327, 331, perception, 10, 307; and phonology, 201;
334; and surface representations, 327; and plasticity of, 276-78; record of impulses,
three-dimensional objects, 330; and the 30; representational level, 39; and sensory
Wiirzburg School, 324 processing, 278; and specificity of func-
mentalism, 66, 194 tion, 273-74, 276, 284; and vision, 321; and
mentality, primitive vs. modem, 232-33 visual cortex, 273-74; and visual process-
Mental Models (Johnson-Laird), 367 ing systems, 319; see also neuroscience
mental models, 332, 364-65, 367-70, 384; neural: bases, 261; cells, 299; networks, 144,
and syllogistic reasoning, 368 169-70, 321; zones, 261
mental representations, 6, 38--40, 128-30, neural functioning, 165-67; and F. J. Gall,
313, 330, 332-34, 368--69, 383-84, 391; 266-67; and the Greeks, 265-66; and local-
and cognitive science, 38--40, 383, 391; ization, 265
and philosophy, 50 neurobehaviorial analysis, and Karl Lashley,
mental structures, 383 13
Method of Serial Reproduction, 115 neurobiology, 388
methodological solipsism, 84 neurochemistry, 282
methodology, molar vs. molecular, 135 neuroembryology, 270
mind: and computers, 6, 10, 51, 57, 67, 82, neurology, 75
148, 166, 223-25, 238--39, 264, 320, 331, neuronal model, 18--19; and logic, 361; and
353; and logical empiricism, 67; primitive, Warren McCulloch, 18, 19
223-25, 232; rational, 57; see also brain neurons, and parallel processing, 318-22
420
Subject Index
neurophysiology, 13, 75, 76, 176, 272, 273- ence, 308, 310, 313, 315; and intentional-
74, 282, 307, 358; and David Marr, 298 ity, 316; and language, 306; in David
neuropsychology, 22-23,260, 272; and Don- Marr's sketches, 301, 303, 307-8, 311; mel-
ald Hebb, 26; and holism, 269 ody, 111; and mental representations, 313,
neuroscience, 54, 260-88; and ablation, 261, 315; movement, 111; and nature of the
273; and behavior, 261, 262, 265; and brain, 300; and nerve-cell sensitivity, 296;
birds, 280-82; and cognitive science, 42, and parallel processing, 318-22; and per-
44--45, 286; and the Hixon Symposium, ceptual array, 311-12; and perceptual psy-
272-73; and holography, 282-84; and lan- chology, 309; and philosophy, 295-96;
guage, 287; and learning, 261-63; and le- principles of, 112, 114; proximity, 114; and
sion technique, 261; and localization, 261, psychology,24-25,295,296,300,302,309,
263, 265, 266, 267-71; and mental imagery, 310, 315; questions of, 295-96; and recog-
330; and molar perspective, 274-78; and nition, 298,304,308,311-12, 321-22;and
rationalism, 379; and reductionism, 261, stereopsis, 302; and symmetry, 114; theory
263, 264, 268, 269, 282, 285, 286; see also of, 286, 307; and top-down knowledge,
holism, localization, nervous system, neu- 304, 308, 318; visual, 131, 169-71, 219,
ral functioning 271, 297, 299; and visual processing,
New Science (Sdmza NOUR), 4, 5 301-5, 308, 319
New York Times, 110 PERCEPTRON, 169
Nobel Prize, 273, 275 phase sequences, 271
phase structure, 187-88, 211, 218; grammars,
186
phenomenology, 286
philologists, 198
philosophy, 31, 49-88, 136, 215-16, 378, 379,
Oedipus myth, 239, 240 381-82, 389-90; agenda, 53; and body, 51;
organizing schemas, 125-6 categories of, 340, 341, 351-5; of cognitive
Origin of Spedes, The, (Darwin), 392 science, 43, 44, 87-88, 389; and computers,
41, 78; critics of, 71; confrontation of Pop-
per and Wittgenstein, 76-77; and empiri-
cism, 52, 54, 60, 86, 206; foundational, 56-
59; and functionalism, 53; and human
thought, 50; and inference, 62; and infor-
paradigms, 130 mation processing, 51; and intentional sys-
parallel-processing systems, 318-22; and ar- tems, 79-81; and introspection, 50, 61, 73;
tificial intelligence, 319; and computers, and knowledge, 52; and language, 62-63,
319, 320-21; and the Establishment, 321; 212, 215-16, 217-20; and logic, 60-61, 360;
and Gibsonians, 321; and MPPS, 319; and and logical empiricism, 60-71; and man as
nervous system, 319 rational being, 360, 369, 378; and mathe-
Paris Neurological Society, 268 matics, 60-61; and mental imagery, 50,
parole, 196, 199 323-24, 329, 330, 336; and mental repre-
Particular Program, see Alfred P. Sloan Foun- sentations, 50; and the mind, 51; mind-
dation body problem, 31, 51, 78-79; and percep-
perception, 51, 100-1, 111-12, 131, 132, 164, tion, 51, 295; and physics, 87; and laws of
169, 170, 294, 295-322, 331, 358; afford- probability, 373-74; and psychology, 61;
ances, 310, 312, 315; and algorithm, 299- puzzles, 76; and rationalists, 53, 60, 86; ra-
300; and artificial intelligence, 169-71, 298, tionalists vs. empiricists, 8; and scientific
304-5, 306, 319; and bottom-up analysis, investigation, 53; and sensory experiences,
305, 312; brain, 112, 307; and categoriza- 52; and systemizers, 77; and the Vienna
tion, 354-55, 358, 359; and color naming, Circle, 63-65; see also empiricism, episte-
349-50; and computational theory, 299; in mology, foundational philosophy; logical
computers, 297-98, 300, 307; and connec- empiricism
tion with brain and computer, 306; direct, Philosophy and the Mirror of Nature (Rorty), 71,
304, 311; ecological approach, 314-15, 316, 76,85
321; and effectivities, 315; and empiricism, phonemes, 26, 200-1, 236; see 11/so utterances
55, 309; Establishment, 312-18, 319, 321; phonology, 188, 200-2, 206, 207-8; see also
Fodor-Pylyshyn view, 312-13; in Gestalt Prague School
psychology, 111-12, 276, 296; goals of, phrase-structure grammar, 186, 187, 188
302; and the Greeks, 296; and illusions, phrenology, 266
254; and implementation, 299; and infer- physicalism, and logical empiricism, 63
421
Subject Index
physical symbol system, 150 quasi-pictorial, 327, 331, 334, 336
physics, and psychologists, 102, 310
physiology, 103; of color vision, 347
plasticity, 262, 276-78, 284; and brain dam-
age, 276; Roger Sperry's results, 276
positivists, 65 Rand Corporation, 25, 145-46
pragmatics, 220 Ratio Club, 25, 26
Prague School, 200-2, 235-36; phonology, rationalism, 8, 53, 81, 86, 291, 317, 360, 363;
200-2 and molecular analysis, 97; and the Vienna
Primitive Culture (Tyler), 227 Circle, 65
Prindpia Mathemafica (Whitehead and Rus- rationality, 360-80, 399; and artificial intelli-
sell), 143, 145, 146, 147 gence, 379; in children, 377; and cognition,
probability, 372, 373, 375-76; and Bayes' 370-79; and computers, 379-80; empirical
law, 374-76 work in, 361; and the Greeks, 360; and
problem solving, in Gestalt psychology, 113 linguistics, 379; and neuroscience, 379; and
procedural representation, 161-62 philosophy, 360, 369, 373-79; and prefer-
process: encoding,131; execution, 131; match, ences, 373; and laws of probability, 372,
131; performance, 131; retrieval, 131 373, 374; and problem solving, 361; and
production system, 131, 150; in Newell- psychological paradigms, 377; and repre-
Simon's model, 131 sentativeness, 372; and laws of statistics,
prototype, 346 374; see also logic, mental models, reasoning
psycholinguistics, 214-15, 217, 219, 220, 379 reasoning, 292, 361-67, 367-70; set also logic,
psychology,10-11,24,25,32,54,69,74,89- rationality
137, 192-93, 194, 198, 201, 215-16, 286, reductionism, 261, 262-65, 268, 269, 282,
310, 324-25, 331, 336-39; and anthropol- 285, 286
ogy, 230, 254; and artificial intelligence, religions, 229
135, 137, 180-81; behaviorism, 109-11, representations: declarative, 161-62; mental,
119, 130, 135, 206, 260; Broadbent's flow 82-83, 330; primal sketch, 301, 303, 304,
chart, 92-93; and categories, 341, 350, 354; 308; procedural, 161-62; 3-D sketch, 301,
and classification, 93-95; and cognitive 303, 304; 2%-D sketch, 301-2, 303; su also
science, 136; and computer models, 95, mental representations
328, 329; and computers, 118-24; develop- representativeness, 372
mental, 116-18, 370; ecological, 315; edu- robots, 141, 173
cational, 370; empirical, 101, 103; and epis- Rockefeller Foundation, 138
temology, 74-75, 85; and foundational Russian Scandal, 115
philosophy, 59; and functionalism, 107-9;
funding in, 25; future of, 137; history of,
99; inference, 100; and information proc-
essing, 29, 90, 91, 119-21, 130-32, 134,
311; and information theory, 95; and in- schemas, 58--59, 116, 118, 126-27, 130, 163,
trospection, 101, 102-3, 107, 119; and lan- 317, 383
guage, 104; and learning theories, 94-95; Science Research Council (Great Britain),
limitations of processing, 90, 95; and lin- 164
guistics, 192-93, 201; and memory, 121- script, 165
24, 126-28, 131; and mental imagery, 323- semantics, 84, 188,191, 205,206,207-8,209,
25, 327, 328, 329, 331, 336-39; and mental 213, 215, 216, 251; generative, 209-10; in-
representations, 95, 128-30, 337; method- terpretive, 209; levels of analysis, 209; and
ology, 96-98, 135, 136; and modular anal- syntax, 212
ysis, 132-33; and molecular analysis, 119; semiotics, 245
and nee-grammarians, 198; and neuro- sensory experiences, 52, 58
science, 278; and optics, 310; and percep- sentences, 63-64, 69-70, 124-25, 182-85;
tion, 24, 25, 100-1, 295, 296, 311; percep- grammatical, 186; kernel, 187, 207; proper-
tual, 134, 309; and philosophy, 61; and ties of, 182-85
physics, 102, 310; and physiology, 103; of SHRDLU, 158-60, 164, 165, 170, .2.97
preferences, 373; and science, 91; scientific, Sloan Foundation, see Alfred P. Sloan Foun-
98-105; and self, 96; and top-down analy- dation
sis, 124-28, 135; see also cognitive psychol- social reality, 235
ogy, Gestalt psychology Society of Fellows (Harvard University),
psychophysics, 96 25-26
psychophysiology, 75 Society of Minds, 166
422
Subject Index
split brain, 275 TOTE (Test-Operate-Test-Exit), 33, 264
S-structure, 218 Tradalus (Wittgenstein), 62
Standard Theory, in linguistics, 207-8 transformational: anthropology, 242.-43; com-
State of the Art Report (SOAP), 36-37 ponent, 208, 210, 217; grammar, 186-88,
stereopsis, 302 215, 216, 218; linguistics, 177, 191, 208,
stereoscopic vision, 300; computation of, 2.51; power, 210; psychological reality,
300 214; revolution, 204, 207; rules, 210-11;
stories, 125-27; and information processing, structures, 210
125-27; schemas, 130 Turing Machine, 17, 18, 31, 79, 144,
structural anthropology, 235, 236-44, 258; 172
and kinship, 236-37, 239; and linguistics,
236, 237
structuralism, 208, 213, 236-44; German,
113; and Gestalt psychology, 113; and
linguistics, 191, 195, 201, 202, 208, 213,
236 Umwt!lt, 31
Structure of LanguJlge, The (Fodor and Katz), 34 unconscious inference, 100
Study of Thinking, A (Bruner, Goodnow, and understanding system, 168
Austin), 93, 95 universal grammar (U.G.), 211-12; subsys-
superordinate, 346 tems of, 211
surface structure, 208, 209, 211 utterances, 69-70, 160, 184, 203; and philos-
syllogisms, 363-64,365-66, 367-70; and men- ophy, 69-70
tal models, 367-70
"Symbolic Analysis of Relay and Switching
Circuits, A" (Shannon), 144
symbolic processes, 243-44
Symposium for Information Theory, MIT,
28 verification, 63
Syntactic Structures (Chomsky), 182, 184, 207, Vienna Circle, 63-65, 70, 73, 76, 292, 351;
213; responses to, 189-90 goals of, 63; members of, 63; Nelson
syntax, 183-84, 185, 190, 191, 193, 201, 203, Goodman's views of, 64-65
206-7, 209, 210, 213, 220, 222; analysis of, vision: and David Marr, 2.98; and neurons,
203, 222; autonomous, 220; and finite- 300; and steps of processing, 301-4;
state grammar, 185, 188; and logical em- stereoscopic, 300
piricism, 64; and perception, 306; and se- visual cortex, 273-74
mantics, 212; theory, 218; and trans- visual perception, see perception
formations, 188 visual processing, 301-5, 308, 319
visual systems, 270
423